Okay, today is Friday, June 25, 2010, exactly 14 days since I took the Miller Analogies Test on Friday, June 11, 2010. My test results are:

Scaled score = 506. Range of scaled scores (they do not provide raw scores, which would be total # correct out of 100) runs from 200 to 600, with a mean of 400, and a standard deviation of 25 points. So, I scored a little past 4 standard deviations.

I AM NOT A HAPPY CAMPER! Although my score puts me in the top 1% of people who have taken this, for all intents and purposes, verbal IQ test, my score does NOT rise to the “Karavitis Standard”!

Although I could easily apply for entry into any of the high-IQ societies with this score, I do not feel that it represents what I am capable of. NO FUCKING WAY!

I intend on re-taking this exam by the end of this year.

(**EDIT: **

**(1) ** Here is a screenshot of the score results:

https://johnvkaravitis.wordpress.com/2010/07/31/mat-miller-analogies-test-result/

**(2) ** An MAT score of 506 translates to an IQ of 168.

**(3) ** For those of you who are wondering, **the MAT is a better predictor of graduate school success than EITHER the GRE exam OR one’s undergraduate GPA.** I have a link to a Powerpoint presentation that was made where the researcher discusses this.

**(4) ** Here’s a link to an alleged conversion from scaled score to raw score.

http://www.csustan.edu/advstd/mat.html

GOOD LUCK!)

Google search phrases: John V. Karavitis Miller Analogies Test, scaled score conversion millers analogy, miller analogies test diagnostic, mat test results, mat test scores and raw score, percentile rankings on miller analogies, mat score percentile, “miller analogies test” scaled score per, mat test score conversion, how many correct answers to get a 400 on, mat raw score percentile, miller analogies raw score conversions, miller analogies test and 90 percentile, scaled scores and correct answers mat, miller analogies test gifted, miller analogy test high scores, average mat score, statistics mat test, mat scaled score conversion, miller analogies test what does my score, miller analogies test score percentile, what is the scoring range on the miller, scale score to percentile rank and miller analogies test, the median score is 400, with a standard, what is the highest and the lowest score, miller analogies test percentile ranks, what does my miller analogy score mean, millers analogy practice convert to mat, miller analogies test score conversion, convert scale score to raw score the mat, miller scaled score, mat scaled score conversion percentile, miller analogies percentiles in population, miller analogies test scale raw to percentile, what does the mat test look like, miller analogies scoring, miller analogies test conversion, how many question right is a raw score of, changes to the miller analogies test, miller analogies test 506, miller’s analogies test admissions score, what is a good mat scaled score, how to figure out your mat score on practice exam, mat percentiles, miller analogies test score conversion, what percentile is my mat score, mat raw score percentile, which is better mat or gre, mat scaled score to percentile, what do my miller scores mean, miller analogies test score conversion, how much correct answer gives 90 percent, mat test score percentiles, mat scaled score raw score conversion, millers analogies test new scale, what is a good mat raw score, mat scoring, miller analogies test deviation, miller analogies test score conversion, whats a good mat score, conversion from miller analogy test to gre, miller analogies test deviation, mat test score required, what does the mat score mean, mat scores and percentiles.

Great score. Thinking about taking it myself.

If you can recall any, could you discuss some of the items you found interesting/challenging?

Hi. Actually, I’m forbidden from discussing what I saw on the actual test. Yuo agree not to reveal what you see as a condition of actually being able to take the test. There are two versions, paper and pencil (which I took – you can NOT write in the test booklet – and, actually, it’s more like a large brochure than a booklet, it’s 4 pages long, thick heavy paper) and computerized. I’m guessing on the computerized one you don’t get to go back and review/correct any answers.

Review books for this and every other standardized test exist at a local bookstore near you. I bought one by Barron’s and took the pre-test, just to see what kinds of questions I’d be seeing. There were 7 practice tests, I didn’t bother doing any of them – maybe I should have. I am really surprised I scored as low as I did, I thought I had done better.

I got my results by mail in exactly two weeks. The paper has three parts to it, your scores are in the bottom third, the re-test form is the middle third.

http://www.amazon.com/Barrons-MAT-Miller-Analogies-Test/dp/0764142356

Good luck!

HI John,

Wow, am I grateful I found your website. I really like your commentation about the actual test. I agree, I think it’s a ridiculous assesment of your intelligence. I am actually very upset today. I just got my score and I used the conversion chart that I found on your site. It looks to me that my raw score is about a 20!!! That is so terrible, I cannot believe. I even had a tudor for a month and we reviewed the Barron’ts review test questions together. I would get a solid raw score of 48–for about 5 of the review tests, towards the end of the tutoring. Not great, but my school requires a raw score of 40. SOOO I guess, I have to take it again. Which I am in bewilderment about because I really don’t think I am going to do any better. It’s not my kind of test. Any kind of encouraging words would help! Maria

if you want to improve your score, the Kaplan test prep book is excellent.

Oops, I meant 10 practice tests!

Ha! Oh, that’s a pity. No other standardized test I know of prevents you from discussing it! :(

Would you say the Barron’s tests were of comparable quality and difficulty to the real deal? I assume you took the diagnostic test? How did you score on that?

In high school I remember taking a few tries from a MAT book by a gal named Lerner. Fortunately, back then the MAT was based on sensible scale of 0-100; your scaled score was your raw score. I’m not sure why they decided to switch. The new scale will most certainly lead to misinterpretation due to a psychological bias, where a score of , say, 470 is viewed as appreciably better than one of 480. When in reality for all we know it was the difference between a 84 and a 85 an on the old exam and obviously not statistically significant. This is why the 1600 scale of the GRE is being changed.

Besides the way it used to be scored you had some freakin idea of how many you missed!

Incorrect. I took both the GMAT and the GRE this year also, they have the same non-disclosure agreement that you have to agree on before being allowed to proceed with the tests. You can’t reveal any questions, you can’t do anything to compromise the test. In fact, I think Kaplan was embroiled in a lawsuit many years back because they would hire people to take the GRE test, I think, and memorize/report back questions.

As for the MAT scale, it’s 6 of 1, half dozen of the other. It doesn’t matter what your score is, the percentile matters most. And even though there might not seem to be a big difference between an 84 and an 85, statistically there is. Remember, I got a 506 and that qualified as 99%. That still leaves a lot of distance to a score of 600. So, I really wonder how many I got wrong. A raw score would have been more informative, tho, I agree.

As for the GRE exam, yes, I know that changes are coming after 2010. They are removing their analogies section from the Verbal section (by the way, the GRE’s analogies questions are way more difficult than the MAT exam’s analogies questions, and not culturally biased as the MAT’s) along with making you calculate an answer in the Quantitative section, and giving the answer, instead of having a multiple choice response.

Just as an FYI, the LSAT used to have an analogies section, which they removed, because everyone found the questions too difficult! LOL!

Oh, and the diagnostic test questions were a “bit” easier than the real thing. As for my score, I scored 80% (96 of the 120 questions).

Fascinating. I had no idea the LSAT used to have analogies! And the fact that they were “too hard” only makes me want to get my hands on some of these old test items all the more. (I use to tutor for the SATs and am a bit of a standardized test lover…I find it quite sad the GRE is doing away with analogies.)

I totally agree that the percentile ranking is the important part. (Though, for some reason, ETS and others remain steadfast in their refusal to give us greater granularity and calculate out past the 99%.)

I also agree that there is a large “statistical difference” in, say, an MAT raw score of 97 and 96, insofar as one is much rarer. What I don’t think, however, is that there is a statistically significant difference between scores separated by a single raw point. That is, I think it rather likely that one could vary by a point or two on any given MAT retake simply by chance. A perfect score of 100 is much rarer than a 99. But I don’t think it tells you much about the difference between the two scorers. And at this stratospheric level, you’re also running into ceiling effects.

94 points separates (3.76 sds) a 506 from a 600. That might seem like a lot, but in reality it’s a comparatively small number of questions. Check thou out this: http://www.csustan.edu/advstd/mat.html

If this is accurate, a scaled score >537 is equivalent to a 99. That means that there could be at most a 63 point scaled score difference between a 99 and a 100.

Which is again why this goofy scale score (while mathematically equivalent) is just ripe for misapprehension and misapplication. lol

You’re incorrect. What makes the scores significant, even with a single point difference, is that you are rating one’s performance against the performance of the tens of thousands of people who took an MAT exam from Jan 1 2003 to Dec 31 2007. That’s why there is a significant difference between an 84 and an 85, or a 94 and a 95. The scores of everyone who has taken this or any other exam fall into a normal curve distribution. Thus, it is correct to say that there is a statistically significant difference between one raw score and its immediate neighbor. The raw scores are normed with a mean of 400 and a standard deviation of 25 points. It’s how the math works, not whether it seems “goofy”. What’s really beautiful about the MAT is two-fold: (1) it has been shown to be a better predictor of graduate school performance than either GPA or the GRE exam, and (2) you only need an exam 100 questions long to get a decent grasp of a person’s mental capabilities.

Good luck on the MAT! I wish you well.

Ah, I think perhaps you misunderstand me. By significant I don’t mean important or meaningful; I mean a p<.05. Blind chance dictates that if you're sitting there on item 99 you have a 20% chance of getting one additional item correct. Hence the standard error on any given MAT score is going to be about 7. And differences in a single item can never reach statistically significance.

Amen. Believe me, I'm well aware of the psychometic beauties of the MAT, economy and predictive validity being just two. And any test with analogies is a friend of mine. lol

Thanks for the good wishes! :)

Make sure to keep us posted on your retake.

I'm sure you'll kick butt–again. lol

The phrase “p<0.05" refers to the significance level used to choose between the null and alternate hypotheses. The researcher chooses alpha, a level of significance, typically equal to 0.05, and then looks at where the result lies after the experiment is concluded. If the resulting p-value lies to the extreme of the normal distribution, such that "p<0.05", then the null hypothesis is rejected, with the chance of a Type I error (rejecting the null hypothesis when it is in fact true) equal to 5 percent. It doesn't relate to what we were talking about here, that is, the percentile ranking of a given test score, based on all the people who took the MAT from 1/1/03 to 12/31/07.

Also, "blind chance" has nothing to do with the discussion at hand.

"Standard error" refers to the standard error of the mean, which is the estimate of the standard deviation of the sample mean estimate of the population mean.

http://en.wikipedia.org/wiki/Standard_error_(statistics)

Again, not quite sure why you brought this up.

I don't believe you have a grasp of the statistical terms that you used in your latest post.

Haha, easy! I know what a p value is….And I have a pretty good maths/philosophy of math background. But if I’m wrong perhaps you can show me where I err?

Some research turned up this:

http://psychcorp.pearsonassessments.com/NR/rdonlyres/1A2076F6-2608-421F-8ECA-EA884EBB9288/0/NAGAPPresentation2008.pdf

Slide 35/45 (reliability measures): “Standard error of measurement (SEM): Estimate of the possible amount of error in a test score range, 6.9 to 7.38.”

Turns out it was higher than my back of the envelope calculation. But I was close enough; and a higher value only buttresses my case.

Fundamentally, SEM is a metric of our uncertainty about a value. Is not a unitary concept and has a number of mathematically equivalent formulations for error-based calculations of various sample statistics. Standard error the mean is just one flavor.

Incidentally, blind chance has everything to do with p values and statistical significance. A p value of .05 essentially means that 5% or less of the time what you’ve observed will come about by blind chance. This is, after all, what the null hypothesis is all about…

But, strictly speaking, this is tangential to my broader point,

which comes down to three basic propositions:

1. Blind chance dictates that a correct answer can be achieved on any given MAT question 20% of the time. 2. Discrepancies of a single point can occur 20% of the time by chance. 3. Ergo differences of a single point cannot be statistically significant.

What’s controversial about that?

By analogy, if the MAT were composed of one question we would not consider a difference in scores (1 vs 0) statistically significant either.

We need not talk of p-values here. P-values have their place in talking about hypothesis testing. The whole idea behind hypothesis testing is that we compare the null hypothesis to an alternate hypothesis. That is, you think you have a better explanation for how the world works, then we compare your alternate hypothesis to the null hypothesis, which says that there is no difference between two sample means. Typically, mu-1 = mu-2, or mu-1 minus mu-2 = 0, however you’d like to present the null hypothesis. Alpha, the level of significance, which is the cut-off point for where we decide we’ve sampled from a different population, is typically set at 5%. That is, we’d only see a score that’s so far away from the mean (i.e. p-value < 5%) if the data are indeed coming from a different population, OR we've managed to pick a sample from the population that does not truly represent it. Given the nature of the normal distribution, this kind of weird sample should indeed be rare, thus the low alpha. To truly knock the null hypothesis off its perch, you need to have results that are way to the right of mu-1 – at least 3 standard deviations, or more. And if you're talking blind chance, I can just as easily pick a sample from the population whose mean deviates a fair amount from the population mean, say one, or even two, standard deviations. Significantly different, yet, given standard practice, that's not enough to knock down the null hypothesis. So, here is a good example of "blind chance" being built into the statistics. And I don't need to discuss p-values or alphas if I'm comparing a test score to statistical data that is presumed to be from the population from which the test score was taken. All I need to understand the test score are the underlying distribution (normal), the mean and the standard deviation.

With regards to blind chance, people taking the MAT will look at a given question and think through the problem. If they're stumped, they'll take a random stab at an answer, of course – they are permitted to; but all of these together will only, on average, give them an additional 1/5 of those guesses as correct. If I guess on 5 questions, on average, I should expect to get one additional point added to my overall "raw" score, and it appears as though that's about, on average, an additional 3 points on the scaled score (per the California State Univ. MAT scaled scores conversions webpage you gave earlier). Not much of a help in the grand scheme of things, as far as I can see. In addition, one can even be right for the wrong reasons, and not be aware of it. Technically, we should dock these particular "right for the wrong reasons" answers as incorrect, as they were not arrived at "correctly". So, overall, there is some error in the test, especially if you have someone who guesses on a lot of the questions (and if they are, they shouldn't be taking the test in the first place), or because a particular test batch had a few more "tough" questions than another batch (which again should be rare). The only way out of this would be to tell people to not answer any questions they are not absolutely confident they know the answers to, that is, that they had not thought their way through the question(s), and to flag any questions that they answer that they decided to guess on (or just dock them 20% of a point for each wrong answer). But that would make the test overly complicated; and a good determination of one's abilities, relative to the population of test takers (which is what the test is all about!), can be assessed, even with the way the test is currently administered. That's where statistical methods come into play. All of this "blind chance" is built into the population statistics.

With regards to the use of standard error of measurement in the slides you reference, I believe that the researcher is giving a confidence interval, that is, we're x% certain (x typically = 95) that one's "real" performance lies in a range. (By the way, this +/- 7 points refers to the scaled score, not the raw score, since slides 28 through 33 inclusive all refer to the "MAT Scaled Score". Since raw scores are no longer released, just the scaled scores, that's all the slides are based on.) A single additional correct question seems to add just a couple of additional scaled score points on the low end of the scaled score system; and, on the higher end, as many as 5 more points on the scaled score system (per the California State Univ. MAT scaled scores conversions webpage).

The range exists because, as highly reliable as the MAT is in test/re-test reliability, people never score exactly the same every time they take the test (for reasons mentioned above). That's what slide # 35 refers to. Here's where your "blind chance" is hiding. It's built into the confidence interval. Per the slide: "…suggest satisfactory reliability according to conventional practices." And if you think about it, they're saying that the score you get (raw or scaled, whatever) is pretty accurate, on average, to the tune of having answered or missed just a single additional question! That's not too shabby a result for 100 questions in an hour-long exam!

With regards to comparing a single test score (raw score), say, 85/100, to the normal distribution created by everyone who took the test from 1/1/2003 to 12/31/2007, a raw score of 86 is indeed statistically different from a score of 85 – it is one point higher, and thus that much further away from the population mean. The area under the normal distribution to the right of a score of 86 is less than that to the right of a score of 85. Thus, the scores are indeed statistically different from each other. Just because one can create a confidence interval around a test score does not mean that we ignore a raw score that is a single point away. I score an 85, someone else scores an 86, he wins, even though we know that if we both took the test 100 times each, we'd each see all of our scores clustered around some mean (for now, the best guess is our first test scores) with a spread of +/- perhaps a point or two, depending on whether we've scored on the low end, the middle, or the high end. In graduate school admissions, if the cut-off for admission is 86 raw (or 480 scaled), I lose, even with the argument that the MAT is not 100% reliable, and even with me pointing out that my confidence interval significantly overlaps the confidence interval of the guy who got into the graduate program.

Indeed, significant changes in test scores from tests taken a number of times by the same person are not expected – you hear this all the time for all the standardized tests, the SAT, the ACT, the LSAT, etc.. And if you're not happy with your first test score, and you take the MAT as second time, if the second test score is very different from the first, Pearson will throw out both test scores. And again, this includes all the times people guess at an answer. Again, your blind chance is built into the statistics.

Personally, I think that one of the reasons that the scaled score has replaced the raw score is that you don't want people arguing over a single point difference. 85 versus 86 raw is arguable given that the scores are "only a single point different", whereas 477 versus 480 has a bit more spread between the scores.

As for the test being created so that it's only a single question, this would never happen – it's nonsensical. With regards to testing, the only way to be sure of anything is to test and test again. And there isn't enough information to create meaningful statistics from a test comprised solely of a single question. So, if your version of the MAT were a single question, I'd have to take it 100 times to be sure how good I'd really score, on average; and then I'd have to compare my average performance against the population of test takers, to really be sure.

And we're back where we started from.

Right! P values are of use in hypothesis testing. And here the hypothesis is that the differences in scores separated by a point is due the ability of the test taker. The null hypothesis is that the difference between observed scores separated by a single point is due to chance. And the associated p value is our probabilistic expectation of single point differences occurring by chance.

Yes, from the SEM referenced you can easily derive a confidence interval. (I thought about framing this discussion in those terms earlier, but for brevity’s sake decided against it.) However, as it stands on its own, the SEM means that any given scaled score is going to have 7 points attributed to error, to randomness, to chance, whatever you want to call it. Hence differences in 1 or 2 raw points are within the margin of error, and not statistically significant. This is not to say that there is not necessarily a massive statistical difference between the scores in terms of their relative incidence. You’re absolutely correct to say the scores are statistically different. But it’s like the political polls you often see. Sure, there’s a statistical difference between 45% and 50%, but since it’s within the margin of error it’s not statistically significant.

I could put forward a more modest proposal that you might more readily assent to: :)

Ceteris paribus, intuitively we feel most confident in a test’s discerning ability between two individuals the larger the difference in score between them.

Thus maximal confidence is attained when the difference in scores is 100 (100-0 = 100). Let us say this denotes a confidence of 100 . Then any diminishment of this score differential will decrease our confidence pari passu. Minimal confidence is associated with scores in which the differential is 0.

Ergo, the difference in 99 and 100 is a state of near least confidence in the discernment of ability between individuals.

You bring up good point which heretofore I didn’t appreciate: how much an average test taker resorts to guessing on the MAT. Preciously few people are as gifted as you. So put yourself in Joe Blow’s shoes. It’s going to be like climbing the intellectual equivalent of Everest. Joe is going to become increasingly frustrated as the test progresses and the items become more difficult and the relationships harder to discern. Half of all MAT test takers are scoring below a raw score of 40! They should get approx. 20 just by randomly guessing! Heck, they probably guess on a significant portion of their correct answers. So Joe’s definitely going to be guessing a freaking ton, esp. on the latter half of the test.

And remember, John, this is a self selected, graduate-school bound Joe Blow. Can you imagine how the randomly selected, man on the street would perform? Your score must represent, what, 5+ standard deviations above that mean? LOL!

Now is THAT up to the Karavitis standard?! :)

(1) We are not “most confident” when the difference in relative scores is 100 raw points (i.e. the maximum difference). We can only truly assign confidence intervals around a specific score, based on a chosen percentage, e.g. 95% confident that…..

(2) Your argument goes back to the “blind chance” idea, which, as I explained earlier, is built into the population statistics. Of the tens of thousands of people who took the exam from 1/1/03 to 12/31/07, we will have seen all types of people and all types of test situations. Regardless, all of their activity is built into the population statistics.

(3) The “Karavitis Standard” is very high, very high indeed. (And my score represents only about 4.24 standard deviations from the mean. I was shooting for at least 7.)

(4) http://www.danielsoper.com/statcalc/calc20.aspx

I just got my MAT score back (exactly 1 week and 1 day after I took it, so I beat you!) and scored a respectable 492 scaled score—also at the 99th percentile!!! In fact, if you check out the Triple Nine Society’s website, you “only” need a scaled score of 472 to qualify. In other words, MAT scaled score of 472 = 99.9 percentile. So my 492 is what, 99.91, 99.92, 99.93, 99.94 ??? And your score of over 500 is what, 99.999, 99.9999 ??? You’re kidding about the “disappointment,” right?

CONGRATULATIONS! You done good!

Yes, I am disappointed re my score. I figure that out of the 100 questions that were graded (of the 120 total), I probably missed less than ten.

I have very high standards. You should know that by now.

As for the Z-score: # stan devs = (506 – 400)/25 = 4.24. That translates into 99.998882%. Here’s the website to calc it:

http://www.danielsoper.com/statcalc/calc02.aspx

Best wishes. Keep in touch.

John V. Karavitis

“Although I could easily apply for entry into any of the high-IQ societies with this score. . .”

Correction—you easily qualify for MANY of the high-IQ societies with this fabulous score (Mensa just being the most obvious, but also Triple Nines, etc.). You would NOT, however, qualify for the even more select societies (e.g., Ultranet, Mega, Olympiq) hence your anger, disappointment, hyperbole . . . . Good luck with the retest. Surely someone with your intelligence will be able to figure out how to post a screen shot of your new score. I, for one, want to see it!

One more thought: your wishes/hopes/plans for a 7 std. deviation score (575) seems misplaced, even for someone with incredibly high standards, as you’ve made clear you have. The range of scaled MAT scores for the 2001-2003 normative sample (I couldn’t find data for the more recent norm group), N = 126,000+, was 231-563. Again, is this just hyperbole or do you really expect a 575 on your retake? The Mega score (1/1,000,000) can’t come around more than once in what, 8-10 years? And you’ll have to do the math for me but your 575 is even rarer than that. Anyway, good luck! If anyone can do it, you can.

Congratulations!

I would submit that retesting would be in order, however the test was in my experience very well designed.

I took it 6-12-2008 and scored a 476, and then I crammed for a year and took it a year later 6-12-2009 and got EXACTLY the same damn score!

I would have been happy with anything over 500. This score places you at the 4rth standard deviation (1:30,000)

Hey Bart. Thanks for your comment, I appreciate it. And congrats on getting a high score on the MAT! :thumbup:

I scored 4.24 standard deviations above the mean. Mean MAT score = 400, stan. dev. = 25; therefore, (506-400)/25 = 4.24 stand. dev.

Not bragging or anything, just trying to be accurate. To get the two-tailed p-value:

http://www.danielsoper.com/statcalc/calc20.aspx

Keep in touch!!!

John V. Karavitis

John, any idea as to the highest score recorded since they switched to the scaled score? I keep checking online with various searches, but it’s hard to get good data because porn sites have saturated the web with meta tags including “Miller Analogies”.

BTW, I haven’t given up on improving my score. I’ve got an idea for an interactive engine which uses the leitner system and will include all the various analogies I can find in print in all the test prep books ever published. It seems to me that the biggest factor in the Miller Analogies is the vast domain specific knowledge needed to be able distinguish the nature of the elements presented.

Hey Bart.

The GRE exam, which I took this year, has incredibly difficult analogies, MUCH more difficult than what I saw on the MAT. But, they are getting rid of these after 2010. If you want to “practice” analogies, get study guides for the GRE exam pre-2011.

Also, the LSAT exam used to have analogies, but, from what I heard, all the “law babies” bitched and moaned about how “difficult” the analogies were, so they eventually got rid of them. Again, get your hands on very old LSAT exam prep guides for more analogies to practice on.

Finally, yes, the MAT exam is “tainted” by culture. But it’s accepted as a standardized test for people wanting to go to grad school in America, and, guess what, we live in a country that has Western European culture as its foundation. Thus, the “culture-oriented” aspect of the MAT. But in my humble opinion, it doesn’t really affect the exam that much. I only recall seeing about a half-dozen or so questions on the MAT that absolutely required knowledge of Western European history and culture.

If you do indeed create any kind of training program, software-based, I hope, you’ll have a killer app for people to play with. Hell, I might just try my hand at it, come to think of it! :) Even just practicing analogies, like working on crossword puzzles or doing Sudoku puzzles, stretches one’s brain.

Dude, keep in touch. And, if you have a blog of your own, post the link here. You can create one for free through WordPress, Google Blogspot, etc.

Take care.

John V. Karavitis

P.S. No, I have no idea what the highest score is out there. I’ll try to find out. If I do, I’ll post the answer.

As for porn sites taking “Miller Analogies” as a search term, I mean, like, yeah, what up with that? LOL!

Cheers.

The highest score in the 2001-2003 normative sample cohort (n = 126,082) was a 563.

(cf: http://psychcorp.pearsonassessments.com/NR/rdonlyres/1A2076F6-2608-421F-8ECA-EA884EBB9288/0/NAGAPPresentation2008.pdf)

Right, that’s what I reported in my post back in August.

Hey! I’m studying for my MAT for my Master’s program in Teaching. I am trying to get a rough estimate of how many correct I need in order to get accepted. I know it’s a scaled score, so it’s not an exact number? But is there a formula to get a general idea? The practice tests just show percentile and my school doesn’t look at percentile. I know it’s a low score, but my school only requires a raw score of 387.

The mean is 400, the standard deviation is 25 points. They don’t give the “raw” score anymore, just the scaled score, which ranges from 200 to 600. You say your school only requires a “scaled” (not raw) score of 387, that’s below the average. Have you taken the test before? Also, Barron’s has a review book you can buy that has practice exams. My personal advice is that worrying about how many to “get right” will not help, just take the test and try your best. But if you really want a conversion from the scaled score to a “raw” score out of 100 questions, someone posted a link to some university in California which has the answer to your question, it’s somewhere in the responses to the “Testing” topics here on my blog.

(Edit: Here it is, you would only need to answer 33% of the questions correctly to get your minimum scaled score. http://www.csustan.edu/advstd/mat.html )BEST OF LUCK ON THE TEST! :thumbup:

A little follow-up: I used McGraw-Hill’s MAT study guide to practice for the exam. There are 10 full-length practice tests in the guide; I took 3-4 of them, in 25-question installments over several days, then tired of the whole process. I was averaging about 77 correct out of 100 questions. (The actual test, as earlier reported, has 120 questions, only 100 of which are scored). On the actual test this past July I received a scaled score of 492 which I believe is in the neighborhood of a 90 out of 100 questions raw score. Needless to say, I scored much better on the actual test than I’d been scoring on the practice tests. Glad I stopped studying ;)

Hi! I am very impressed with your score. I have just graduated from college and need to take the MAT for admissions to graduate school. I’m required to get a total raw score of 30. Can you tell me approximately out of the 100 questions how many I would need to get right to get that score?

Hi Katy.

A scaled score of 383 translates into a raw score of 30 right out of 100. I posted the link to the alleged conversion up above, but here it is again.

http://www.csustan.edu/advstd/mat.html

Get your hands on the Barrons MAT review book, it has a pre-test and 10 practice test. Schedule yourself where you are taking one of these every week up until your scheduled exam.

http://www.amazon.com/Barrons-MAT-Miller-Analogies-Test/dp/0764142356

Good luck!

Katy,

To get a raw score of 30 you need to answer correctly . . . 30 questions. That’s the idea of “raw score,” i.e., number of correct responses. The only catch is that the MAT has 120 questions, not 100, because 20 of the questions are “experimental,” i.e., they’re being field tested for use in future and will not make up part of your raw (or scaled) score. So, to be certain (in a probabilistic sense) of a raw score of 30, you will need to answer 50 correct. In other words, if you answer exactly 30 correct out of 120 but 20 of those correct answer are on the “experimental” questions that aren’t being scored (highly unlikely, I know, but just in case), you’ve only really got a raw score of 10. Answer 50 correct out of 120, take away the 20 experimental questions, and you’re left with 30. In this scenario, probability suggests you’ll have more than 30 correct but we’re using the “worst case” scenario to make sure you get at least a raw score of 30. Does this make sense? Regardless, I’m not sure I’d recommend your going into the test with a “I need this minimum raw score to pass” mentality. Better to have prepared (not “crammed” or even “studied”) to the best of your ability, weeks or months ahead, and to go in with a “I’m going to do my best” or even “I’m going to kick a#%” attitude and then see what happens. Good luck!

great score! I’ve read that your kind of score is very rare! you should be proud. I am also planing to take the MAT this coming Sunday for my grad school application packet, my school requires a score of 396 to get into their grad program. I know that it may looks easy to get that score, but I am gonna aim high towards getting a better score than that. :)

Good luck reaching your goal score!

Hi Richi. Many thanks for your kind words!

A scaled score of 396 translates into a raw score of 38 right out of 100. I posted the link to the alleged conversion up above, but here it is again.

http://www.csustan.edu/advstd/mat.html

Get your hands on the Barrons MAT review book, it has a pre-test and 10 practice tests. Schedule yourself where you are taking one of these every week up until your scheduled exam.

http://www.amazon.com/Barrons-MAT-Miller-Analogies-Test/dp/0764142356

Good luck!

FWIW, I took the MAT in 1998 trying to get into a M.Ed. program. Back then they reported the raw score but not percentile rank. The necessary minimum raw score they wanted for unconditional admission was 36 out of 100. Pathetic, I know. I got a 78 which upon later review I found out is about the 98%ile. So, it seems there is plenty of fine detail in the test. I am not a humanities/lit type of person, so I was surprised to do even that well. I figured I could get above 36 on any test just because that is such a low score, so I didn’t do any prep or anything. I literally didn’t know what at all to expect other than the obvious that it would be analogies.

As for analogies being hard. I don’t get that. If you don’t know what one of the choices is or one of the parts of the incomplete analogies, then, yeah that makes it hard to pick the right answer. But that is just plain knowledge deficit, which was my problem. If you know what all of the things are, then picking the right one isn’t so hard.

A 78 was higher than the 98%. By comparison, I believe a scaled score 66 on the old MAT was the MENSA threshold (which is 98% on IQ or standardized test batteries). Mind you, there use to be some eight different versions of the MAT of somewhat varying difficulty; form J being the most difficult, if memory serves. But I’m certain that a 78>98% on all of them.

I just could not depart your website prior to suggesting that I extremely enjoyed the standard info a person provide for your visitors? Is gonna be back often to check up on new posts

Ineed to score in that top 40th percentile. Can you tell me what that score needs to be?

The population mean score is 400, the population standard deviation is 25. To score in the top 40%, you would need to score a 406. z-score = (Test score minus population average) divided by the standard deviation. A z-score of 0.2534 would mean a test score of 406, or, a score better than 60% of those taking the exam. Here is a link that gives you a picture of the bell curve and the calculation: http://danielsoper.com/statcalc3/calc.aspx?id=2

So, that’s just a little bit over an average score. With a little bit of effort, that should be attainable. Good luck!

John V. Karavitis

Thank you so much for your response. I really appreciate it. What time frame do you suggest to study/prepare for this exam?

Hi John, your score sounds very impressive.

So I have a question. I took the GRE twice and did horrible both times. I took a kaplan prep course and it still didn’t seem to help. I don’t know if I’ll succeed in the MAT if I did so bad on the GRE. Would you recommend taking the MAT? I’m scared that I’ll do just as poorly and I’ll be even more depressed. Thanks for any advice you can give.

My reply is that it never hurts to try. You’ve already assumed the worst, why not study up for the exam with Barron’s exam prep book, available through Amazon.com. Then, give it a shot. The GRE is a bit more involved and tests thinking ability slightly differently, the MAT only lasts an hour at most. Good luck!

John V. Karavitis

Hello, John!

I hope that you’ve been well. Firstly, I wish to congratulate you on attaining a score of such an elevated calibre! I imagine that you must enjoy approaching intellectual challenges in order to have procured such a standing, and I am therefore glad to correspond with you, as I too have always found it, erm, fun to pursue high results on interesting assessments, and to venture into areas of academia that I hadn’t previously explored.

With that said, I will beg for your aid. I’m contemplating taking the M.A.T. relatively soon; I am aiming to achieve a score of 500 or more, albeit that I would like to strive for as high as score as I may attain. I doubt that I will be able to do it, as I’ve always thought myself to naturally lack intelligence, but I love academia and have braved certain tests with moderate success (triple 800s on the S.A.T. this year, if that’s at all relevant, and a childhood I.Q. standing that would permit me entrance into, if I recall correctly, at least the Prometheus Society). To be brief, then, whilst I don’t believe in myself, I will try my best to make this work!

On that note, could you please advise me as to which preparatory texts proved the most useful? I haven’t previously utilized prep. books to study for standardized tests, and I therefore don’t know. I’ve found material from Barron’s to accumulate good reviews; what do you think? Furthermore, could you elaborate upon any correlation between the M.A.T. and the S.A.T. that you may have stumbled upon? I speak not of the content of the tests – I am aware that they differ markedly on those grounds. I’m inquiring more as to skills that one might be able to extract from one’s S.A.T. experience and apply to the M.A.T.

Lastly, to be terribly vague: tips, please? How was your M.A.T. session? Would you advise studying for the test, or simply having the standard “good-sleep-and-good-breakfast” combination and going for it? Which aspect of it did you find challenging? Specifically, any tips regarding the paper test? The testing center I’ve chosen does not offer the computerized version.

Thanking you most sincerely,

Sophia

Hi. Thank you for visiting my WordPress blog, and for your kind words.

With regards to “studying” for the MAT exam, such a thing is not really possible. What I recommend is getting a hold of a copy of the Barron’s MAT Review book, available from Amazon.com. It has a total of 11 practice tests. I recommend taking one every two weeks in a quiet setting, like a library on a Saturday morning, until you feel you are ready for the real thing.

The reason that you can’t really “study” for the MAT exam is that it is one of the best measures of true intelligence. Analogies are at the heart of manipulating ideas. This is something that you can’t “game”, but you can, however, sharpen your test-taking skills and reduce your test anxiety. The MAT exam is supposed to help people to, in part, decide whether they are capable of tackling a graduate degree. Thus, success on the exam will not only depend on one’s native intelligence, but also on how well one has applied oneself in school. The MAT has a very high test/re-test reliability factor, probably the highest of any of the “intelligence” tests. Your score will more than likely change very little on a second or third attempt.

As for the actual test-taking scenario, I briefly touched upon this in an earlier post. The exam that I took was proctored by a small university located in downtown Chicago. The exam was one hour long, and scheduled to start at 5:30 PM on a Friday afternoon. I almost came a half hour late to the exam, I mistakenly thought that the exam started at 6 PM! The exam is a stiff cardboard 4-page brochure-type booklet, and besides instructions on the front page, all it has are 120 analogy questions, exactly like those found in the Barron’s test review book. (Some locations have the exam on computer – ask before signing up. You may prefer the pencil and paper version, with that, at least, you get to go back and review if you have time. I finished in about half an hour, and went back and pored over the ones I thought I had some trouble with. I don’t think you get that opportunity with the computerized version.)

If you’ve scored 800s on the SAT, I would think that you would score well enough on the MAT to get into any graduate program you would want. Take a few practice tests, and relax. Getting all worked up right before the test will hurt more than anything else.

Best of luck, please keep in touch and tell us how well you score.

John V. Karavitis

I do want to mention that I just took the computerized MAT today and you can go back and recheck questions, or skip questions to go back to later. After question 120, you have the option to ‘recheck all answers’ or ‘recheck skipped questions’ to either go back to 1 and head through again or go back to the first one you left blank and continue through in order. The one thing you can’t do is skip around easily.

Hey Jeremie. Thank you very much for that info. I know that they still offer both the written and computerized exams. Someone posted a comment a while back, they confirm what you have just reported.

It’s unfortunate that they don’t make it a little easier to maneuver one’s way around the exam, hopefully they will address that issue in the future.

May I ask, were you given your score immediately after the test? Is that possible? Or do you have to wait for your test score in the mail? If you indeed got your results, please share them with us.

One thing that they might try in the future is making the exam “computer adaptive”, as the GRE and GMAT exams, that is, they start off with a question of average difficulty. If you answer it correctly, you get a tougher question, if not, an easier one. Might be able to take the test down from one hour to maybe, what, 15 minutes? How would THAT be for what is essentially an intelligence test!

Best wishes in your future endeavors.

John V. Karavitis

I got my “unofficial preliminary score” about 30 seconds after I hit finish – 451. I’d hoped for higher, but expected lower. I decided to take the MAT 7 days ago, and powered through the entire Barron’s book in that time, just doing practice tests. My weakness on the practice tests – which I think was also the case on the actual exam – was vocab and recognizing science/social science references to people and their work, not the relationships. I do wonder if I could have increased my score a bit by being a bit more methodical about my prep, since it was a subject matter deficit.

I did find it a much more relaxed experience than the GRE (2006 – 710V, 650M) which I gave myself a month of prep for and spent that time almost exclusively relearning the math, with some vocab building flashcards between classes.

I’ll admit I was a bit relieved when I found out the MAT wasn’t computer adaptive, because I do like to skip around and the inability to do so on the GRE was a time management / testing strategy difficulty for me.

When did you take the GRE? I took the GRE in 2004 and in 2009 (or thereabouts), they used to have an Analogies section, but I believe that in 2011 they took it out. I think the GRE is a\ lot different, too, in that you have to give them an answer, especially for the Math section – no more multiple choice.

As for your MAT score, I think you did great. The scores are normally distributed. The mean is set at 400, the standard deviation is 25, and you scored 451. Okay, so, rough guesstimate, that’s two standard deviations above the mean.

https://en.wikipedia.org/wiki/Standard_deviation for a cumulative distribution percentile of 97.8%, which is excellent. By the following link, it’s exactly

97.932484%http://danielsoper.com/statcalc3/calc.aspx?id=53

Either via the GRE or the MAT, you did well. BE PROUD!!!

John V. Karavitis

My GRE was Fall 2005, so they’d expired by the time I needed scores for a new graduate program. (I mistakenly put 2006 in my other post… but that was my grad year). I wasn’t eager to take the new GRE (or revisit the math again), so I opted for the MAT.

Is there a chart for converting MAT scores to their IQ equivalents? Thanks much! -Kat

Good question Kat. No, not to my knowledge. However, what I did was assume that performance on the MAT translates into and reflects general intelligence. On a typical IQ test, the “average” IQ is scaled at 100, with a standard deviation of about 16 points. My score on the MAT exam is 4.24 standard deviations from the mean. Therefore, assuming that intelligence is in general reflected by performance on the MAT exam, and that in the general population intelligence is normally distributed, you can convert an MAT score to an “IQ equivalent” quite easily. I’m guessing that this is also what “high IQ” societies or groups do when they try to determine what an “acceptable” score is for membership. These groups have a number of different IQ tests that they will consider, and they must use cut-offs that reflect very high percentiles.

Again, this is a ballpark estimate. Also keep in mind that we have relatively very few people with standardized test scores that high, so it becomes increasingly difficult and problematic to say just what the difference is between someone who has an IQ of 160 vs. someone with a 170, or a 180, etc. I’m guessing that my performance on the MAT means that I missed less than ten questions (see link http://www.csustan.edu/advstd/mat.html ). How does that compare to someone who only missed eight? Or five? Or even none at all? And so on. John V. Karavitis

Since most people who take the MAT are college graduates, the mean IQ of this group will be higher than 100. MENSA requires a 95% on the MAT which they say puts one in the top 2% of the general population. Therefore, the conversion, as you state it, will not be accurate. I scored a 471 on the MAT which would correspond with a 145 IQ based on your method. But since the mean of the test takers was most likely more than 100, my IQ would be greater than 145.

Incorrect. Your assumption that college graduates, as a group, have a higher mean IQ than the normal population, is false. The MAT qualifies as an intelligence test, and it has a normal distribution, with mean 400, standard deviation of 25, and a range of 200 to 600. If you would have taken the time to click on the California State Univ @ Stanislau link that I provide, you would have seen that people do score terribly on the MAT, even given their college education and practice exams, like Barron’s et alia. The assumption that high-IQ societies use to look at different types of IQ tests to determine qualification for membership is that there is a general ability called intelligence, people’s scores fall into a bell-curve pattern, and that these tests do not require any specialized training to understand and work the problems. Thus, to compare performance on one IQ test to another, you’d focus on how many standard deviations above the mean one scores.

It’s that simple. Sorry to hear you couldn’t pull off a full 3 standard deviations when you took the MAT. But thanks for playing.

Hello John,

I took the MAT today and got a preliminary score of 402. I have been searching ways to convert this number into the percentile range. My number one school requires a score in the 50th percentile.

I know my 402 isn’t as attractive as your score. I admire your level of intelligence and would appreciate any help you are willing to give. Thanks and Congrats!

The MAT is scaled on a normal curve, with

the mean at 400, the standard deviation at 25 points, and the range from 200 to 600. You literally just scored a hair’s breadth over the 50 percentile mark. Daniel Soper has links that allow you to convert scores to percentiles, etc.http://www.danielsoper.com/statcalc3/calc.aspx?id=22 Your score of 402 translates to a z-score of 0.080.

http://www.danielsoper.com/statcalc3/calc.aspx?id=53 Your score of 402 translates into a cumulative distribution of 0.53188137, or the 53rd percentile.

Best of luck to you in your future endeavors.

John V. Karavitis

Not sure you will ever see this Sunshine, but what was your final official scaled score? Was it still 402?

Just a quick note, since you mention it several times in your posts: I took a computer-based test today (at University of Illinois-Chicago’s Testing Center), and you are in fact able to return to skipped questions, and to review the whole thing.

When you reach the end of the 120 questions, the computer gives you three options: returning to the ones you skipped, reviewing all, or completing the exam. I finished in 40 minutes, skipping three or four; after choosing an answer for those, I went back and reviewed all my answers. The one disadvantage is that you can’t “jump” to a specific question, the way you can in a physical book–you have to toggle through, using the “next” and “previous” buttons.

Page 38 of this PDF shows a screenshot of the computer: http://www.pearsonassessments.com/NR/rdonlyres/1A2076F6-2608-421F-8ECA-EA884EBB9288/0/NAGAPPresentation2008.pdf

Thank you for clarifying this. I have become very wary of paper and pencil tests, I think the LSAT is the only exam which still stubbornly refuses to go electronic. Electronically-administered tests are better than paper and pencil tests, no chance of any accidents, monkey-business or cheating occurring. And with electronic tests, your results should be available immediately, and you should be able to schedule the exam at your convenience, just like scheduling the GMAT or the GRE through Prometrics.

The set-up that you described appears to make it no different than taking the paper and pencil tests, then. My main concern with that was being able to go back to test items that were a bit too difficult, and, of course, you’re not penalized for going back to review your responses and double-checking. Good to know.

Best of luck on your exam results and your future endeavors. Tell us how you score, we’d like to know!

John V. Karavitis

The instant score I got was 463, which surprised me. I expected to do better, although I have to admit I didn’t really prep at all. I’m sure it was the math-related analogies that hurt me; I haven’t used a natural log in 20 years, and so have lost all knowledge I once had of how they function.

I’ve already been accepted to my master’s program; the scholarships I’m applying for require the MAT. {shrug} We’ll see what happens.

What range of scores do they hint that they want to see? Or do they not provide that info? If not, can you get that info and post it here? I’m sure the university admissions advisors might be able to give you an idea of a minimum score that might be needed for these scholarships. I would be interesting to see how many standard deviations above the mean they want to see.

Good luck.

John V. Karavitis

I got my final results in the mail today. My scaled score of 463 did not change, but I was in the 99th percentile for the total group, and 98th percentile for my intended major (which is Library Science, if I didn’t say it before). I returned to complete an undergrad program as an adult and have a 4.0 in that (so far…still have 4 weeks to go), so hopefully that bodes well for scholarships.

It is interesting to me that even with the 40 point difference in our scaled score, we both came out in the 99th percentile…I’d love to see the decimal points: was I a 99.2% and you a 99.8? Does anyone ever score 600? 550?

Hi. That’s cool. I wouldn’t have thought that the computer would have changed your score. You did great!!!

The MAT has the highest test-re-test reliability factor of any of the standardized achievement/IQ tests. It’s a rather accurate assessment of a particular aspect of one’s intelligence and educational achievement. It also has greater predictive value for success in a graduate program than either one’s GPA or GRE exam score. http://psychcorp.pearsonassessments.com/NR/rdonlyres/1A2076F6-2608-421F-8ECA-EA884EBB9288/0/NAGAPPresentation2008.pdf

The powerpoint presentation that I have linked to in my thread says that 563 out of 600 is the highest recorded MAT score. A score of 463 is less than three standard deviations (MAT mean score = 400, stan dev = 25). Daniel Soper’s link calcs that a score of 463 is 99.413226%; a score of 506 is 99.998882%. To put it in context, a score of 463 or better is seen once in every 170 test attempts, whereas a score of 506 or better is seen once every 89,445 attempts. Remember, the MAT is scored on a normal distribution.

http://www.danielsoper.com/statcalc3/calc.aspx?id=53 Mean = 400; stan dev = 25

Is a perfect score possible? Certainly. But beyond a certain point, I think that the score carries less meaning, since, obviously, we have fewer and fewer people scoring at such a high level. In addition, as you can see from http://www.csustan.edu/advstd/mat.html , at the top end, we’re talking about missing just a handful of questions. How much difference is there between missing 10 questions, vs. 8? Or just 6? Not a lot. So, the farther out you go, the less meaningful the score is, although very impressive.

I appreciate that you kept us up to date on your efforts, and I congratulate you on your performance. Best of luck in your future studies.

John V. Karavitis

I took the MAT (computerized version) a few days ago. I studied for a week and a half, each day taking practice tests and reviewing the rationales. I used a book by ARCO and found it to be very helpful. My preliminary score was 407. The proctor at the testing center told me that my score could increase with my official score due to how the weight of the 20 experimental questions affects the scored test questions. Sounds weird…is that even possible???

My score barely met my school’s minimum requirement of a 403. I feel sick about it. Wondering if preparing a bit longer and taking it again would substantially improve my score???

Thanks for any advice or input.

Hi DMichelle. Good job on taking the MAT, it’s a deceptively simple exam that humbles many many people who take it.

You scored a 407, which is a little bit over the average (see earlier comments in this thread re statistics et alia). What I have read is that the 20 experimental questions do not in any way affect or impact your test results. They are created to give the test creators an idea of how difficult potential new test questions are. Now, I do not have anything to do with the creation, distribution, administration or grading of these tests, but what I have read, esp. in the Barron’s exam prep book (available on Amazon.com), is that the 20 experimental Qs do NOT affect your grade. (Edit: Well, since you don’t know which ones are the experimental Qs, in a sense they waste your time, but they do not affect your score directly.)

You passed your target school’s minimum score, so effectively you are done with the MAT. You could re-take it, but, as noted in a Pearson powerpoint presentation noted earlier in this thread, the MAT has one of the highest, if not THE highest, test-re-test reliability factor of any standardized IQ-type test. That is, is is highly unlikely that re-taking the MAT would materially change your test score. You also have to keep in mind that (1) a lower subsequent score could sway the Admissions committee negatively and (2) if you re-take the MAT within one year from your last test, you have to give the test proctor that third part of your mailed test score that tells them what test to NOT give you. At least, that’s how it works with the pencil and paper version of the exam, the computerized version may know automatically based on your Soc Sec # etc.

If you do want to re-take the exam, get your hands on the Barron’s review book, do one practice test a week, and in the interim try doing crossword puzzles. The whole idea is to sharpen your MAT test-taking skills, there is no way to “beat” or cheat or “game” the test. All you can do is become more familiar with how the test will challenge you, and to sharpen your test-taking skills and be more confident going in.

Best of luck, keep us advised as to how things turn out for you.

John V. Karavitis

Addition: A score of 403 translates into a percentile of 54.78%; 407 is 61.03%. You did better than you think. Mean = 400, stan dev. = 25.

http://www.danielsoper.com/statcalc3/calc.aspx?id=53

Also, it only took a couple of questions answered correctly.

http://www.csustan.edu/advstd/mat.html

Thanks John. I will keep you guys posted if I decide to retake the exam in the future. Now that I’ve done a little research, it does seem that I did better than I originally thought.

DMichelle, was your final official scaled score also 407 as well? If not, what was it?

Congratulations on the 506, John! I am (tentatively) thinking of taking the MAT in the spring of 2012, and am therefore trying to figure out just how difficult I ought to expect the exam to be. How would you say these questions compare with those on the MAT? (I didn’t include problem set #1 because its questions are tragically easy.)

http://www.matexampracticetests.com/practice2.html

http://www.matexampracticetests.com/practice3.html

http://www.matexampracticetests.com/practice4.html

Lena C. Ruiz

About average for the exam. There are a number of questions on the actual exam that are real head-scratchers, tho. I recommend getting a copy of the Barron’s MAT Review/Prep book off of Amazon.com, and taking an exam at the library under test conditions (quiet, well-lit table). I encourage you to take the actual exam, just practice once a week with a practice exam (there are 10 in the Barron’s book), try doing crossword puzzles on a regular basis, and then take it. You’d be surprised how few questions need to be answered to get just an average score. Look here: http://www.csustan.edu/advstd/mat.html 41 questions correct out of 100 is a score of 400, which is the average score (50th percentile).

I wouldn’t recommend taking a review course/class unless you take the actual exam and do terribly. I don’t recommend a review course for two reasons: (1) what will they teach you that the Barron’s book won’t?, and (2) why spend money on an exam that you really can’t “beat”. I recommend taking the MAT in the afternoon, not first thing in the morning. I also recommend the computerized version, not the paper and pencil.

Best of luck and tell us how you do!

John V. Karavitis

DId you ever re-take this test, as you mentioned you would, above? If so then what were the results? If not, then do you plan to re-take it? Given your own statements about the test’s reliability, that would seem futile, if your aim was to reach a significantly higher score; perhaps you realized this?

Hi Kevin. No, I never re-took this test. I’ve got my fingers in a lot of pies, so to speak, which keep me really busy (e.g. running races and reading and writing book reviews on Amazon, etc.), and also, given the high test re-test reliability of the MAT, I sort of put any plans to re-take the MAT by the wayside. I have really high expectations when I set out to do something, I guess I sort of took my test results a bit too seriously. As you can see from http://www.csustan.edu/advstd/mat.html , I didn’t really miss that many questions, probably no more than 10 out of the 100.

Your comment comes at an interesting time, though, since I was indeed thinking about this very thing just a few hours ago. (Synchronicity, ka-ching!) I think I definitely will take the test again sometime in the near future, perhaps I should do so on an annual or bi-annual basis, if for no other reason than to prove to myself that my brain is still in “good working order”. It would also be interesting to see how my performance would decline over the succeeding years. (Not by much, I hope!!!)

Take care.

John V. Karavitis

I just got back my MAT score. I scored in the 97th percentile in the total group and intended major. I really have a hard time believing that this is an accurate gauge of real intelligence. So much of the analogies I had required trivial knowledge. For example, one was something like Harpers : Ferry :: Valley : (forge, cove, something, something). I sat there staring at this for like a minute trying to figure out what the heck it was talking about. I then just decided to put Forge because somewhere in the recesses of my mind Valley Forge sounded correct. When I got home to look this up, it turns out Harpers Ferry and Valley Forge are the names of two national historical parks. Are you kidding me??? I get IQ points for knowing (actually, guessing!) that?!? Another one I had was Apple : Cider :: Pomegranate : (juice, grenadine, etc…) I just have a hard time accepting this to be intelligent thought.

One thing, for anyone who is taking this test, should know is that the math analogies can be quite harder than the ones seen in the Barron’s or Kaplan MAT books. I’ve perused both for a couple weeks before the test. The math analogies on either required no more than several seconds of thought. For the actual MAT, I had analogies that required a minute or more of my time. One actually had me rechecking my work and the answers for the right thing. I was looking for an algebraic form that was similar to the one used. It turns out, that was a trick. The ones that were similar in form to the analogy were “wrong” and not the expanded algebraic expression. Instead, the algebraic expression that was the right analogy was in a completely different form but equivalent to its analogy. I know this isn’t making much sense without a concrete example. There was another that said (x^2)^1/2 : 3 :: (algebraic expressions) : 27. So, just realize the math portions of the MAT could be much harder than the simple 3^2 : 3^3 :: 9 : 27.

Hello David. Congratulations on scoring so high, and also in doing well enough for your future academic plans Also, thank you for sharing your experiences with us.

I know that some of the questions might seem dumb or ridiculous, but remember, that particular question may have been one of the 20 experimental questions, not one of the 100 that are used to grade your performance. I’m not sure how the test creators decide which experimental questions are useful or good, maybe the “Valley Forge” one is a dud.

I know that the MAT has been criticized for having a “cultural” aspect to some of the questions. Nevertheless, you are taking an exam whose score will be used to determine whether you will enter a graduate program in the United States. In addition, the few questions on the exam that have a cultural bent to them will not make or break you. There are not that many of these types of questions, you can still do well and get a score good enough for your needs.

If you don’t believe that analogies are really as hard as people claim, or that they are a true measure of that nebulous thing called “intelligence”, then I refer you to versions of the GRE exam pre-2012. The GRE recently dropped its analogies section. I believe they did so because so many people found the questions to be so difficult, even more so than the ones seen in the MAT. Don’t believe me? Get your hands on a GRE review book pre-2012, and see how hard their analogies questions are. They do NOT use any cultural questions, they are strictly by-the-book words from the dictionary. In fact, I believe that the LSAT used to include an analogies section, but as with the GRE exam, no more.

David, you did well, be proud of your achievement. And as a parting thought, although you found some of the questions easy or nonsensical, many many test takers have a hard time with them. If it were that easy for everyone, it wouldn’t be useful as a screening test.

Best wishes in your future endeavors.

John V. Karavitis

Hi, John.

Thanks for the reply.

I suspect that one reason for the high reliability of the MAT is that it tests crystallized intelligence. My understanding is that, unlike fluid intelligence, crystallized intelligence often increases with age and experience. For that reason, I’d be surprised if your subsequent performance were to vary much, either upward or downward, until advanced age or organic disease set in. On the other hand, since you are an outlier at the extreme right tail of the distribution, you might see a bit of regression to the mean, but nothing extraordinary, I would imagine.

Just for fun, I think that I am going to take the MAT later this year, myself. And if you want to tackle some really brutal analogies questions, by the way, then see Lewis Terman’s Concept Mastery Test, Form A! (Read more at the following link, if you are curious: http://hiqnews.megafoundation.org/2002-9-5_More_on_IQ.htm).

Hi Kevin.

Thanks for your thoughts and the link. Good luck on the next test session.

John V. Karavitis

Hi John,

Thanks for the response. I don’t mean to antagonize anyone on here, and if I sound like I am, I apologize, but the idea of IQ testing and intelligence has been one that has bothered me quite a bit for the past 13 years. I think I have enough anecdotal information for me to formulate a new hypothesis. It is one that is shared by Marilyn Vos Savant, whom most people interested in things like IQ know about: it is ridiculous to measure intelligence with just a number. Hear me out.

First, I scored really poorly on the first SAT test way back in 1992. I think my initial score was just over 800. I failed for many reasons, but mostly it was because I was a very poor student. I had to bulk up my rusty math skills and took the test again several months later. I ended up scoring 1020, or something. In terms of IQ correlation, this shouldn’t sit well with people. It makes no sense to me that something innate like intelligence can be boosted that high with some amount of studying. After college, I took the GRE and scored very high. In the quantitative, I scored in the 99th percentile. My IQ correlation was boosted even further since high school. This is just the beginning of many odd discrepancies I’ve noticed with IQ tests. The latest of which is probably the most damning.

[SECTION DELETED]

I could go on and on about weird discrepancies I’ve noticed like this over the years. Richard Feynman is a stark example, and so is Gary Kasparov. Infact, Ronald Hoeflin, the founder of several high iq societies also admitted that his iq based on various tests have ranged from 125 to 175 depending on the test. He claims his actual IQ to be around 160s. What I’ve finally come to realize, and this is really liberating to me, is that one test does not measure a person’s intelligence. Intelligence is so varied that I don’t think most standardized tests can accurately capture it. I have recently seen the WAIS testing methodology, and I think it’s probably better at creating an overall picture of IQ or intelligence. It tests memory skills, abstract reasoning, math skills, etc… However, it is not without flaws. There are serious cultural problems with it. In my abnormal psychology class, the professor told us one of the questions they use as a determination of intelligence: you notice smoke and fire in a theatre; there are people in there; no one else notices; how do you proceed? The answer to get 3 IQ points was retarded. Even he, and the rest of the class, agreed.

Sorry for the long diatribe of a post, John. I just had to finally get this off my chest. Hopefully someone finds it useful in this quest called life. My overall message to that person is don’t worry about scoring low on a standardized test. It doesn’t mean you’re not intelligent. If Feynman, with a score of 125 IQ in high school, can become a Putnam Fellow (this makes Feynman a math genius) and Nobel Laureate in physics, you can at least finish grad school and become successful in your own right. Don’t put much stock in “IQ tests”. I don’t.

Hey David.

No offense taken. I appreciate your thoughts.

[SECTION DELETED] As for standardized tests and what they mean in the grand scheme of things, they have their place. In general, smarter (and more educated) people will score better than those who are not as smart and/or not as educated. They are not perfect, and they test a narrow range of analytical skills. (Although the MAT has been statistically shown to be a better indicator of graduate school performance than either one’s GPA or the GRE exam.) And also, with so many people applying to graduate programs, there has to be some fair way of applying a cut-off to all the applicants. Typically, schools use a formula that takes both the MAT score and one’s undergraduate GPA into account, and these formulas are derived from linear regression, so they are based on actual performance of enrolled graduate students.

Keep in touch, tell us how you’re doing.

John V. Karavitis

Tell me what you mean by the word “intelligence” first, before you look at someone’s standardized score and deem that person “smart” or “not so smart.” Talk about circular reasoning. “Smart people will score better than not-so-smart people.” So who are the smart people who will score better on the MAT? The ones who got good grades in school? The ones who got high scores on the SAT? Guess what grades in school measure? All too often they measure one’s ability to take a test. Or one’s ability to conform to the teacher’s expectations. Standardized tests “have their place in the grand scheme of things.” OK, I’ll stipulate to that; but that’s not a definition of “intelligence,” either. That just means that schools want a simple, fast, quick way to do some sorting. The scores are a better predictor of one’s likely success in grad school? Again, that’s not a definition of intelligence, only another way of saying “We’ve found someone who’s good at school.” The scores “test a narrow range of analytical skills.” Yes, thank you. But who says those narrow, analytical skills are the end all/be all of what it means to be intelligent? At least acknowledge that these tests measure a small slice of what it means to be intelligent. And in this day and age, a not-so-special slice. Commentators all over the place are talking and writing about the need for students with more right-brained skills than the more typical left-brain skills that schools have emphasized for eons. (See Daniel Pink’s A Whole New Mind) And these “tests of smartness” are left-brained tests. Surely you’ve heard of Howard Gardner’s theory of multiple intelligences or Tufts University Prof. Robert Sternberg’s triarchic theory of intelligence, no? Intelligence can broadly be defined as the ability to learn, to pose problems and to solve problems. Gardner talks about ALL the ways that humans can be said to be intelligent and each of the several intelligences that he’s identified is just as important as all the others (Logical-Mathematical and Verbal are only two of the many human intelligences yet they’re the only two that these vaunted tests—SAT, GRE, MAT, IQ—purport to measure). Prof. Dylan Evans makes a case for yet another intelligence: the ability to assess risk, probabilistically, and use that sense of the likelihood of events in your everyday life. Intelligence surely includes higher order, creative thinking skills, the ability to think divergently and convergently, and the ability to bring metacognitive skills to the table. Tell me how a 1-hr. multiple choice test measures anything of the sort.

Riddle me this: Who’s smarter, someone who scores 500 on the MAT or someone who can take a car apart (or your house’s plumbing or electrical work, etc.) and put it back together? The one who scores 500 or can paint the Mona Lisa? The one who scores 500 or composes the 5th symphony? Me (492 on the MAT) or Mike Holmes, the construction genius on TV? And if Mike Holmes took the MAT and scored 450? Does that matter? What does it say about intelligence (or about your view of it, more precisely) that you think it can be reduced to a score or percentile or stanine or standard deviation? I submit that intelligence isn’t found in a paper and pencil (or computerized) test. It exists in the world when you take your unique skills and talents out into it and make your way there. It’s visible only at the intersection of the human mind and the world it navigates. You’re not intelligent because you’re a good test-taker (or else your intelligence is very very very limited and not particularly useful to society). The fact that your high scores can help you gain entrance to college is just an artifact of a broken (overwhelmed) system of higher education. I’m not sure that a 500 on the MAT says any more about your intelligence than an ability to type 90 wpm does. At least with your typing score, you have some job prospects and can offer society a valued skill and work product!

John D. I hate to say this, but it has to be said:

“You sound upset”.

The MAT exam has been statistically demonstrated to be a better indicator of success in graduate school than either one’s undergraduate GPA or performance on the GRE exam. It most definitely is a solid measure of a person’s intelligence, and it is incredible that you can get a strong idea of someone’s intellectual abilities from a pencil-and-paper test that lasts, at most, a mere one hour. The MAT also has the highest test/re-test relaibility of any standardized intelligence test.

As someone with a life-long interest in the human mind and psychology, in addition to a Bachelors degree in Psychology, I am well-versed in the alternate theories of intelligence that you mention in your harshly critical post. In fact, my final paper in my Cognitive Psychology class dealt with the issue of expert performance, and it noted the work of K. Anders Ericcson at Florida State University. However, we live in the Information Age, and the one facet of intelligence that is most prized in our society today is the analytical. I do not discount or denigrate the other “forms” of intelligence. However, as you yourself must know, other theories of intelligence that seek to “parse” intelligence more finely all agree that there is a fundamental intelligence, or intellectual ability, or “general” intelligence factor, referred to as Spearman’s “g”. In addition, many different types of intelligence tests have been found to be highly correlated with each other, along with the fact that studies have demonstrated a high correlation between ability in one area of Gardner’s areas in his Theory of Multiple Intelligence with others (e.g. mathematics and music).

https://en.wikipedia.org/wiki/Intelligence

I agree with you that it matters more in Life how one applies oneself than just one’s natural talent. Nevertheless, one cannot apply oneself with some natural talent acting as a foundation for all the hard work necessary to be successful in Life. Intelligence has a biological foundation, but it will not flourish without a nurturing, supportive, and challenging environment in which to express itself. As I like to say, you can build a larger house on the stronger, deeper foundation. Nature is the foundation, nurture is the house. You get nowhere without a healthy dose of both.

Thank you for your passionate, albeit somewhat vituperative, posts to my blog. I appreciate and respect your position.

Cheers.

John V. Karavitis

P.S. I can type fast enough, although it’s of the “hunt and peck” variety.

P.P.S. Oh, and I scored a 506 on the MAT.

Hi John,

Again, much thanks for the thoughtful response. I just realized, on my daily run, that I was not being very thoughtful at all in mine. Can you [SECTION DELETED]

Anyways, you the man. Wish you the best in life!

Hi David.

Relevant sections of your posts and mine have been DELETED.

John

Hi. How did you convert your MAT score into IQ? You said your 506 translates to 168 IQ. Just got my scaled score of 451 when I finished the test today. Good enough enough to get me into grad school…. except for my lackluster GPA.

Hi Sarah. Good job on the score, it translates into 97.9 percentile.

As for the conversion to an IQ score, I did it by assuming that the MAT test evaluates intelligence in general. That is, it isn’t a test of how much math you know, or any really specialized knowledge. Rather, it’s a test of a fundamental activity that reflects intelligence: handling analogies. Typical IQ tests that are administered have a normalized standard deviation of 15 to 16 points, with a mean of 100 (the “average” IQ is 100, and the curve is a normal distribution). A score of 506 translates into about 4.24 standard deviations, so, comparing the result on my MAT test and other intelligence tests (which is how high-IQ societies compare tests against each other), then I should be expected to score in the 160s on a general IQ test. It’s a matter of assuming that intelligence is a general ability, and that distinguishing one person’s performance on different tests can be done by looking at how many standard deviations above the norm one scores.

From an earlier post today, here is how you can calculate your score of 451:

The scores are normally distributed. The mean is set at 400, the standard deviation is 25, and you scored 451. Okay, so, rough guesstimate, that’s two standard deviations above the mean.

https://en.wikipedia.org/wiki/Standard_deviation for a cumulative distribution percentile of 97.8%, which is excellent. By the following link, it’s exactly 97.932484%

http://danielsoper.com/statcalc3/calc.aspx?id=53

You did great, be proud!!

John V. Karavitis

Well, I’ve read through most of the comments on this topic, and there is only one clarification needed to end all discussion, one way or another; did you, John V. Karavitis, retake the Miller Analogies test? If so, why have you not posted either a triumphant vindication of your personal estimation of your abilities or, in the interests of evenhanded fairness, post that your retake of the test resulted in a lower or equal score?

I don’t see any need for the rest of the commentary, it’s just dross that can be tossed aside after a resolution to the apparent outrage of the original post is provided.

No, I, John V. Karavitis did NOT re-take the MAT. I’ve been busy with other things. I think I will re-take it in the near future, and I promise that those test results WILL be posted here.

Remember, the MAT has the highest test-re-test reliability factor of any of the currently available IQ-type tests.

Thanks for your concern, Alex. Keep in touch.

John V. Karavitis

Thank you for the reply.

By the way, the old Stanford-Binet IV scale which used 16 points per Standard deviation, is no longer used extensively. The Stanford Binet V, like the WAIS-IV scale uses 15 points per Standard deviation; 4.24 sigma is an approximate deviation IQ of 164 by the most current standards. Any approximate correlation with results correlated from older tests would provide a higher IQ number, which is no longer valid, objectively speaking. By the standards of thirty years ago, a current 164 would be, in most cases, equivalent to ~174.

I think you need to step back from yourself and get some perspective.

But good luck with it all.

https://en.wikipedia.org/wiki/Intelligence_quotient

Although current tests have a stan dev of 15 points, older tests used 16 points. From the Wikipedia article:

“The values of 100 and 15 were chosen in order to get somewhat similar scores as in the older type of test. Likely as a part of the rivalry between the Binet and the Wechsler, the Binet until 2003 chose to have 16 for one SD, causing considerable confusion. Today almost all tests use 15 for one SD. Modern scores are sometimes referred to as “deviation IQs,” while older method age-specific scores are referred to as “ratio IQs.””

So, 15 or 16 points, okay, but you get the idea: I kicked ass on the MAT. How did YOU do, by the way?

So, thanks for your concern. I’ll file it in the circular file, as it deserves. Now, run along and go back to worrying about the environment, and all that.

Oh I see, so now you, “kicked ass” rather than outrageously failed due to some kind of testing error or some other perceived handicap. Come on.

I understand your need to destabilize an uncomfortable situation with aggression, it’s a good way to throw someone off balance, shock them into silence, or push them into making an error. It’s a common trait.

I haven’t taken the MAT yet, but since personal experience taking the test plays no role in what I’ve said, it’s not relevant; just a possible ad hominem hook, on your part, to proceed with more aggressive defensiveness.

I’ll leave you with the last word, since you’ll sit there and wait and wait until you get it anyway.

It’s okay to cry. Thanks for playing. YOU are the weakest link. Good-bye!

JOHN

CONGRATS ON YOUR SCORE.I ONLY GRADUATED HIGH SCHOOL BUT IN MY LATER YEARS-40 TO PRESENT-HAVE MADE A HOBBY OF PSYCHOMETRICS.WHEN THE MEGA TEST HIT OMNI MAGAZINE,I SCORED A 25. 18 OF THESE WERE VERBAL. I LATER FOUND OUT THAT THE VERBAL QUESTIONS WERE REDUCED TO I/2 POINT FOR EACH! OH WELL..

SINCE MY SPATIAL ABILITY IS SO POOR-I CONCENTRATED ON THE MAT. I TRIED BARRON’S PRACTICE TESTS AND OF THE FIVE I TOOK-MY LOWEST WAS 64 AND HIGHEST WAS 68. I CAME TO YOUR SITE TO GET AN APPROXIMATE CORRELATION WITH THESE SCORES AND I.Q. RANGE.

WHAT ARE YOUR THOUGHTS? THANKS IN ADVANCE!

PS-EXCUSE THE CAPS LOCK-I DO NOT SEE SO GOOD.. (:

05-12-2012

Hello Robert.

Thank you for your kind words.

To convert the scores that you earned on the Barron’s practice test, I would like to suggest that you look at the number of questions answered out of the 100 practice questions, and compare it to the percentages found in the California State University – Stanislaus website.

http://www.csustan.edu/advstd/mat.html

68 correct answers out of 100 translates into a current MAT score of 441. The mean for the MAT is 400 (normal distribution), with a standard deviation of 25 points. A 441 would translate into cumulative probability of 0.94949742, or, in terms of percentiles, 94.9-th percentile.

http://danielsoper.com/statcalc3/calc.aspx?id=53

With regards to IQ, it depends on which assumption you make. Typically, IQ is also distributed in a normal distribution, with a mean of 100 and a standard deviation of 15 (or 16, for the older tests) points.

https://en.wikipedia.org/wiki/Intelligence_quotient

You scored 41/25 = 1.64 standard deviations above the mean, so 1.64 * 15 + 100 = 124.6 IQ.

Take care.

John V. Karavitis

HI JOHN,

THANKS FOR YOUR RESPONSE AND THANKS ALSO FOR THE LINKS.I REALLY APPRECIATE

YOUR TIME!

Wow, what an interesting thread to read! I am a Kindergarten teacher heading back to school for my Masters. I got a 4.0 in undergrad. Took the MAT today and was poking around to see how my preliminary score of 461 stacked up. I was curious because the administrator of the test seemed pretty impressed with it. I hopefully used the calculator you linked to get the percentile and it spat out 99th? Really?? If I did that well, dang. Cool. Looks like this Kindergarten teacher needs to join MENSA! Did I calculate that percentile right?

Hello Carol.

Good job on your score. Congratulations.

Had you read this thread closely, you would have come across

http://danielsoper.com/statcalc3/calc.aspx?id=53

MAT is normally-distributed, mean = 400, standard deviation = 25. Therefore, a 461 ==>

Cumulative distribution function: 0.99265637

or, percentile-wise: 99.265%

Best of luck in your future endeavors.

John V. Karavitis

John, thank you for the insight and discussion. I’m not sure you can help but I had a basic question.

My girlfriend just took the MAT today and her preliminary score was 408. I know that is not the official scaled score but can you explain how much the preliminary score might decrease/increase?

For example, your official MAT score was 506. Was your preliminary score also 506? Greater? Less than?

I only ask because she is applying to grad schools which require at least the average 400 score for consideration and I would like to know if her 408 score is “safe”. Thanks in advance for your time.

Hello Robert.

Sorry, I have no idea. Although I recommend that everyone take the computerized version of the MAT, I took the paper and pencil version. There was no preliminary score for me, just a wait of exactly 14 days before I got my results in the mail.

I think the “preliminary” score is given to give the test taker their score, but it is subject to review, that is, if they find that they scored the exam incorrectly, or if there was a glitch in transmitting her answers to the testing mainframe, etc.

The MAT exam has the highest test/re-test reliability of any standardized test. On a second or third sitting, I doubt that anyone’s score would change significantly.

As to whether a score is “safe”, I understand where you’re coming from, but exam scores are one factor in grad school admissions. Grad schools look at past data, which includes undergrad college, GPA, exam scores, etc., and they create a linear regression equation. This equation is then used to create a minimum “entry score” for admissions considerations. It’s all very mechanical, and serves two purposes: (1) it takes the guesswork out of the admissions process, and (2) takes the human bias factor out, which helps protect the grad school against charges of racism, sexism, etc.

Your girlfriend scored above the average of 400, so she’s above average, of course. She should contact the admissions dept. and ask how low of a MAT score the school typically considers. Even if her score is below their threshold, other factors (GPA, undergrad college, etc.) may swing things in her favor.

http://danielsoper.com/statcalc3/calc.aspx?id=53

Normal curve distribution, mean = 400, stan dev = 25, score = 408 —-> Cumulative distribution function: 0.62551583

Per this website, your girlfriend scored in the 63rd percentile. That is, out of 100 test takers, she scored better than 63 people taking the exam.

Best wishes to you and your girlfriend, and take a moment to remind her how lucky she is to have someone like you care and worry about her!

Take care.

John V. Karavitis

Thanks for the response but you misunderstood what I meant by “safe”. I of course understand that there are multiple factors in terms of admission to a particular grad school. What I meant by safe, was that her preliminary score of 408 would remain at, or very near, 408, and especially above the 400-point threshold. Since you did not take the computerized version, you obviously cannot answer with any authority, though I appreciate the reply.

I doubt that the “unofficial preliminary score” will change. Although I did not take the computerized version, a prior poster in this thread, Jeremie, did, and he said that he got his “unofficial preliminary score” 30 seconds after hit hit FINISHED. If his score had changed at all, he would have said something. SO, as I presume, they want to have time to make sure there was nothing wrong with any of the questions, that there wasn’t any snafu with the servers, etc.

Take care.

John V. Karavitis

That was what I assumed but I thought it would be worth doublechecking. Apparently, people actually called Pearson to ask the very same question and were told the chances of the preliminary score changing at all was close to nil. Thanks again though for your thoughts.

Im trying to sort through all this and just want some help….I need a score of 400 on the MAT. I am taking the practice tests from a book and need to know about how many questions I need to get right. Can you please help me! Thanks!

Use this link as your guide:

http://www.csustan.edu/advstd/mat.html

Best of luck!

John V. Karavitis

i have a questions that i reallly need answered. if i took a test with 25 questions and i know i got 5 wrong what is my score PLEASE ANSWER THIS

Your “score” would be 80%, no? 20/25 right = 0.80

However, your question is ill-formed, and more than likely meant to be a question re similar performance on the MAT. For that, look at http://www.csustan.edu/advstd/mat.html where 80 correct out of 100 = 464 on the MAT.

Cheers.

John V. Karavitis

Just as an FYI., It seems that the idea is to repeat tests so one can score higher and that would translate to a higher SD above the norm and equivalently higher IQ. Hate to disappoint. You’re fooling yourself and corrupting the real test. You cannot do that. Additionally, look at the standardization and realize that an SD of 25 has far more sample size round the mean than at extremes. Effectively there is a far greater standard error and as you deviate above 3 or 3.5SD, scores become so varied that exact estimates are far less accurate.

Funny that a “psychometric expert on exceptional intelligence” would make such claims. I never made any of the assumptions that you state. FYI, sample standard deviation is calculated based on the sample, and the normal distribution assumption and resulting sample statistics are assumed to represent the population of test takers. If you were the “expert” that you claim you are, (1) you wouldn’t post anonymously, (2) you’d understand the difference between a standard deviation and a standard error (check out http://en.wikipedia.org/wiki/Standard_error and get your terms straight), (3) you’d know that the MAT has the highest test/re-test reliability of any of the standardized tests that have ever been created. FYI, your statement “Additionally, look at the standardization and realize that an SD of 25 has far more sample size round the mean than at extremes” is all the proof I need re your real employment status.

I’ll grant you that things get “fuzzy” the further out toward the tails-ends of the normal distribution, but then again, the underlying assumption is that these test scores are distributed in a normal distribution, that the standard deviation captures the spread, with all that entails.

As for taking the test again, scoring higher, and this translating to a higher IQ score, given the reliability of the MAT test, this may not be possible, I will grant you that. However, one can achieve a higher score by practicing the skills that the MAT tests, by being less anxious before and during the test, by having had enough sleep the night before, etc. A lot of variables affect one’s score, but then again, all of this should be encapsulated into the sample standard deviation that is calculated from the tens of thousands of people who have taken this test over a period of a few years. (You’re the “psychometric expert”, you know all of this, right?)

Thanks for playing, you’ve amused me to no end. (By the way, what was YOUR score on the MAT test? What’s that? You never took it? I wonder why….)

John after reading all the post, yes all of them I am feeling less anxious about taking the upcoming MAT. I purchased the Barron’s study book after looking at all the other study guides out there and it seem to fit my needs.

Let me ask you do you think the practice test on the Pearson website is a good indicator of how well a person will do on the actually test?

P.S. you are my HERO in the world of TESTING!!!

Don’t ever feel anxious about any standardized test. A test only gives an indication, it isn’t an absolute, or the end of the world if things go wrong. Although the MAT is an excellent indicator of success in graduate school (applying to graduate school is typically the only reason that it’s taken), it is only one factor. Different schools have different cut-offs, etc. And you also have the right to take the test again, of course.

I never took any of the practice tests on the Pearson website. You say you have a copy of the Barron’s practice test book, I recommend you take one practice test every weekend, in a quiet environment, like a college library. It will give you an idea of what to expect, and also give you a bit of practice in dealing with analogies.

Thank you for your kind words. I wish you the best of luck.

John V. Karavitis

I am following up! Thank you so much for this BLOG, It was so very HELPFUL!!!! And thanks for your response. I took the MAT yesterday and I am happy to announce I made 400!!! Whewww, I took your suggestions and it worked. THANK YOU SO VERY MUCH and best of luck to you, I am on my way to to apply for Grad School!

Congratulations on your score! Extremely impressive!! What is the minimum score required to meet a 40th percentile?

I believe it would be a score of 394. Using the cumulative distribution function for the normal distribution, along with a mean of 400 and a standard deviation of 25,

http://www.danielsoper.com/statcalc3/calc.aspx?id=53

Cumulative distribution function: 0.40516513

Obviously a score of 400 would be the mean, and thus the 50th percentile.

A 394 would mean getting 37 questions correct:

http://www.csustan.edu/advstd/mat.html

Good luck!

Dear John,

On April 18, Joshua Walker wrote:

“Since most people who take the MAT are college graduates, the mean IQ of this group will be higher than 100.”

You responded:

“Incorrect. Your assumption that college graduates, as a group, have a higher mean IQ than the normal population, is false.”

This is wrong. College graduates, as a group, do have a higher mean IQ than the general population (about 1 SD above the population mean). Information about this can be found in the technical manuals of many standardized IQ tests. You can probably also find this information in “The Bell Curve” by Herrstein and Murray. If you can’t find this information, do let me know; I’ll forward you a few relevant papers published in scientific journals.

Michael

Hello Michael.

I have 6 university degrees. From personal experience over a span of a quarter century, I assure you, most college graduates have no business ever having gone to college in the first place. The same holds true for people who pursue a Master’s level degree, I have two.

Since the college degree is now the new high school diploma, the only qualification that one needs to get into college is a pulse, and to graduate, the ability to warm a seat for four years. College graduates as a group are as dumb as the average person because, at this stage of the same, they ARE in general the average person. The studies which you have most graciously offered to direct me to are testing the incidental advantages of book learning, which is unavoidable when you spend four years with your butt glued to a seat. Most people who are in college, about 70%, are there to better their job prospects, not because they are smarter than average and have a desire to learn.

As for people who take the MAT, or GRE, or GMAT exams, etc., they, as a sub-group of college graduates, may indeed have a higher than average IQ. But college graduates in general are really dumb. Not to play the “paper wars” game, but studies also show that the average college graduate can’t write, even to put together a coherent resume, can’t think logically, etc.

Thank you for viewing my WordPress blog, and your thoughtful comment.

John

Hi John,

As “dumb” as the average college graduate may be, he/she is still brighter (or less dumb, if you prefer) than a person with an IQ at the mean of the general population. Data pertaining to this can be found for all subtests of the Standford-Binet, Wechsler scales, the Cattell Culture Fair Intelligence test(s) and the Raven’s Progressive Matrices, to name a few. It’s not simply book learning. In fact, even for simple processing speed, the average college graduate will tend to outperform a person with an IQ at the population mean.

You wrote “…studies also show that the average college graduate can’t write, even to put together a coherent resume, can’t think logically”.

Correct. You may be shocked but things are even worse for a person with an IQ at the population mean…

Michael

Do you have any plans to join Prometheus? Looks like they’re accepting new members now and your score definitely meets the criteria. I do have to say that they strike me as insufferable (while you most certainly do not). As you and others have noted, there are different types of intelligence and a lot of the traditionally gifted/talented folk seem to be lacking in regards to emotional intelligence..

I am a “triple niner” but I find myself wanting to try for more. I know that it doesn’t make a lot of sense from a practical standpoint and I’ve never participated in any of the societies anyway. There’s always that little voice, and I think you may know which one I’m talking about! Anyway, that’s how I found your site. I was quite pleasantly surprised and am glad I stopped by. :-)

Your most recent post is interesting – about the fact that our education system is designed so that any mindless automaton can complete the requirements. My teenage daughter who is a 98 percent-er uses the same rationale to explain her lack of motivation to do as well as she could. My thinking is that it is an expensive pre-test for the real world because it’s easy compared to life. You are right – it has become a basic prerequisite that for the most part doesn’t really prepare people for the working world unless they plan to enter academia.

I believe that the fact that a lot of people come out of college not knowing how to write well and do other things (rational thinking), could be laziness or lack of training rather than a lack of ability or potential. A lot of people (myself included) aren’t living up to their full potential – and that’s a continual challenge and struggle throughout your entire life. To be better.

My question is – are they really dumb (as you say) or are they just behaving that way because they always have and it’s easier? “Stupid is as stupid does”, I guess. Sure, there are people who do their best and will be nothing more than mediocre – but it is believed that nearly anyone could become world-class at anything with 10,000 hours of experience. Some exceptional individuals need far less time, but that’s beside the point. Could you apply this to logical thinking and problem solving?

Hi, Shan, to answer your question the simple answer is that most traditional college students just do not care about their education. I am a non-traditional student attending a community college and that is the general attitude that I perceive from many young people in school; they just want to put forth the minimum effort to get by so that they can pass a set of courses and earn a diploma with the goal of finding a better job. Most of them can really care less about elevating their minds and their characters with challenging ideas. Even some of the professors when asked lament about the general, lackluster attitude of the majority of these students. In my opinion, they come across as woefully dumb because they are so unmotivated to care in the first place.

But there are exceptions out there–and the only real difference between the dumb ones and them is that they really do care about their futures and about learning in order to better themselves as intelligent adults who will have something to offer back to their communities someday.

Hello.

People go to college for a number of reasons. Studies show, however, that 70% of college students are there to improve their future job prospects. Those who are “real thinkers” find themselves at college as a way station toward graduate studies and perhaps a PhD.

One might find it contradictory that someone who complains that “most traditional college students just do not care about their education” would himself be going to a community college.

John

Dear John,

First, congrats on your exceptional MAT scaled score of 506! Now, “as noted in a Pearson powerpoint presentation noted earlier in this thread, the MAT has one of the highest, if not THE highest, test-re-test reliability factor of any standardized IQ-type test. That is, is is highly unlikely that re-taking the MAT would materially change your test score.” If that is true, then why the contradictory assertion at the beginning of this thread:

“Although I could easily apply for entry into any of the high-IQ societies with this score, I do not feel that it represents what I am capable of. NO FUCKING WAY!

I intend on re-taking this exam by the end of this year.” ?

Why even attempt another re-test, when even in your own words when doing so will not materially improve your score??

I just wanted to point this out, as I found the obvious self-contradiction in you interesting.

I have very high standards. Even a high score of 506 leaves me hungry to do more. Granted, given the high test/re-test reliability of the MAT exam, such a thing may not be possible, but I plan on trying again. Not this year, I have too much on my plate, but definitely by the end of 2014.

John

Hi, I have a score of 435. What is the percentile? I enjoyed the test but it was a challenge with the questions about music and wars.

http://danielsoper.com/statcalc3/calc.aspx?id=53

Mean = 400, standard deviation = 25. x = 435.

Cumulative distribution function: 0.91924334

92nd percentile.

You got about 65 of the 100 questions correct:

http://www.csustan.edu/advstd/mat.html

John,

Congratulations, both on your fantastic score and an inspirational blog!

I am humbled by the amount of diverse knowledge that you intrinsically possess to get such a score on the MAT. A true polymath..

I do believe that a certain amount of preparation is possible for the MAT by reading the Barron’s and/or other books. The content in the books can help refresh some rusty concepts / terminology that are lying dormant in the deeper recesses of the mind but you need to know the stuff prior as it usually takes years to accumulate.

The MAT is a mile long and an inch deep and that defines it best.

I gave the MAT a fair try today and I got a 455. I compute this to be in the 98th percentile via the links you’ve provided in your posts.

I see that, Shan, a poster above, incidentally, is an alumnus from IITK, which would mean that he passed the grueling entrance exam to that institute and placed himself in the top 0.1% among the exam takers, which IMHO is an achievement of great magnitude given the deep subject expertise involved. That exam is a mile long and a mile deep!

All the very best in all your endeavors, John!

Hello Ajay.

Thank you for your kind words, and congratulations on your MAT score.

Best of luck in your future endeavors.

John

Hi John,

I need your advice.

I saw a sample MAT score report on a Pearson document to illustrate how the score report actually looked and it is available on the link below:

The percentile scores do not seem to match the scaled score.

Your thoughts?

Agreed, it looks a bit off, but not by much. Given a normal distribution, a mean of 400, a standard deviation of 25, and a score of 410

http://danielsoper.com/statcalc3/calc.aspx?id=53

yields a cumulative distribution of 0.65542174

so, 65.5%-tile.

They are claiming 70-th percentile for the test group. They may be working off of a different batch of test takers (unlikely) or it’s rounding to the next highest 5th-percentile, or it’s human error.

A score of 410 means about 47 questions answered correctly – less than half. Nothing to write home to mom about.

Hi John,

Thanks for your prompt response and the enlightenment contained within.

I did a little digging and I found that the current normative sample are the candidates who took the test for the first time between January 1,2008, and December 31,2011 (n = 125,910), which may include your good self.

http://images.pearsonassessments.com/Images/dotCom/milleranalogies/pdfs/MAT_Reliability-Validity_FNL.pdf

I wonder how this will affect my score and percentile?

I took the test on January 15, 2014. I studied from some free online questions about a week or two before the exam, and received a 447 scaled score. That is about 1.88 standard deviations above the mean. This puts me in the 96th percentile, according to a z-score calculator. That is good enough for many good doctoral programs in the United States, including the education program I am applying to.

I thought most of the analogies were very easy. I spent an average of 8 seconds per analogy. I think most of the analogies I may have missed involved words I’ve never seen and whose relationship was vague to me. Most of these were probably the experimental analogies.

hello, how long did it take to get your official score in the mail?

Exactly two weeks. I took it on a Friday early evening, got the results in the mail two weeks later, on a Friday.

I believe it’s all computerized now, you get a preliminary unofficial score right after you finish the exam, and then they mail you the official score. I took the paper-and-pencil version.

Best of luck.

John V. Karavitis

I received my official scores 10-14 days later.

I have not opened every single link posted here so perhaps I may have missed this already having been established, but have we figured out whether anyone has received a perfect score of 600? I seem to remember a link someone posted above which indicated the highest score being a 560-something. I’d be thrilled to know, and it seems Mr. Karavitis is shooting for the elusive 600.

In the Pearson pdf file (the link doesn’t appear to work, but I downloaded it way back), one of the slides is titled “Distribution of Scaled Scores by Intended Field of Study and Total Group for the 2001–2003 Normative Sample”, and shows a high score of 563 (on slide 40 out of 45). The range of observed test scores was 231 to 563.

And yes, one should always “shoot high,” else what’s a heaven for?

John,

I, finally, managed to track down a document from Pearson that has the scaled score to percentile (total) for the MAT exam based on the latest normative group, applicable as of Jan 1, 2013, based on scores of candidates who took the test for the first time between January 1,2008, and December 31,2011.

I have extracted the relevant information and is available here:

The entire document is available from the Pearson site and is at:

http://images.pearsonassessments.com/Images/dotCom/milleranalogies/pdfs/MAT_Comp_GREFlyer_v4.pdf

A scaled score of 410 DOES correspond to a percentile of 70 on this chart (in regard to my query above) as does other anecdotal MAT scores and percentiles reported in various blogs/posts.

Since your blog IS the most comprehensive source of information on the the actual MAT experience and relevant discussions, I feel the above document has found its rightful place.

Also, the scores of the current normative group do not seem to follow a normal distribution which IMHO, should be the case given the fairly large sample size (n = 125,910), A superficial analysis yields that the MAT percentiles are ahead of the percentiles derived from a normal distribution by approximately 5.

Higher percentiles for the same scaled scores between the different normative populations, 2001 – 2007 vs 2008 – 2011, seem to indicate declining intelligence, quite converse to what the Flynn effect claims.

A possible explanation for this could be the currently omnipresent trend of placing extensive reliance on the gifts of modern technology, to remember information/ facts.

Your thoughts?

http://www.danielsoper.com/statcalc3/calc.aspx?id=53

Mean = 400, standard deviation is 25, score is 410

“Cumulative distribution function: 0.65542174”, or 65.5% percentile.

I’m thinking rounding up. Also, please note, Pearson says that the distribution is normal, with a mean of 400, and standard deviation of 25 points. Yet the document you have referred to and link in your comment has a score of 400 being at the 55-th percentile, directly contradicting Pearson’s assertions re the statistics of the MAT exam. Quite curious, but not earth-shattering or overwhelming.

Thank you for adding to the wealth of knowledge about the MAT exam. I, and future readers of this blog, appreciate your efforts.

John

On further thought, maybe they have the columns for the percentiles of the MAT and GRE switched? By accident? Take a close look at the percentiles…. for MAT score of 410, they have the GRE percentile at 65%

Thanks so much for your answer John! Congratulations on your score. My preliminary score was a 426. I only had about 2 weeks to study. I bought the MAT Kaplan book (kindle edition) on Amazon and also used a few practice quizzes online. The book helped me a lot! The only thing I really did though was study vocabulary words from the huge list in the book, didn’t bother with any of the other stuff – art, music, history mythology, etc. Wish I had because a lot of my questions were on mythology and history. A few of the questions were totally the same as the book. I’m just so relieved it’s over with and I never have to think about these crazy analogies again! I get excited whenever I’m reading a random article and see a vocab word I studied though! This website was pretty helpful for me…200 analogies questions with thorough explanations of the answers (only works in study mode for me): http://www.practicequiz.com/examQuestions.php?ExamId=61&tk=bc99211951e3ba2af658136733000af0

I can understand not being happy with that score. You bested me–I’m “only” in the top five percent (I got a 497), and those Kaplan books led me to believe I’d be scoring in the mid-fives. Pearson’s prep materials led me to expect somewhere in the lower-fives (by which I mean about a 525). I suppose I shouldn’t complain, but it certainly was a bit of ego deflation. Which is probably a good thing, but it’s not what I purchased the test book or downloaded the practice tests for.

I agree with the statement about “most college graduates have no business ever having gone to college in the first place,” but it’s not about intelligence–it’s often about attitude and discipline. there are some people who just aren’t meant for higher ed right out of the high school gate. there are some who aren’t meant for it for decades later. there are some who never are, and that includes some of the purported best and brightest. I’d take a hard worker with an average intelligence and a zeal for learning over a purported genius who doesn’t give a crap about anything, any day.

I took the MAT in February of 2012 and received a score of 451 (and my previous comments on this post can be seen further up-thread). TODAY, April 5, 2014, I received a letter from Pearson dated 3/26 that my exam “was scored incorrectly” and a new score report was issued with a score of 454.

I find it both amusing and distressing that there was a scoring error that took over two years to notice, even if it wasn’t terribly significant.

Hey Jeremie. Nice to hear from you again.

Congrats on your score, it’s quite high, two standard deviations from the mean.

Hope it helps you get where you want to go. Keep us in the loop.

John

Hey! Would you mind if I share your blog with my twitter group?

There’s a lot of folks that I think would really enjoy your content.

Please let me know. Thanks

As long as you mention my full name, John V. Karavitis. Also, I have authored a number of essays in popular culture and philosophy books, please share this link also:

http://www.amazon.com/John-V.-Karavitis/e/B00GA2FRPU

Thank you.

In high school I scored 97th percentile on the School and College Abilities test verbal reasoning portion. I answered 47 out of 50 analogies correctly. Any idea how that translates to score on the Miller Analogies Test?

You need to see how many standard deviations from the mean that score represents. You are trying to compare two different tests, although they both claim to test the same analytical ability. If it’s an “easy” test, it also won’t mean much.

Hi! Sorry to ask when so many others have, but I’m a bit mathematically challenged and would love your input on my MAT score. I took it today and got a 452 preliminary scaled score. What would the approximate percentile be for this (I say approximate not because math can change but because Pearson has the final say…)? I would have tried to calculate that myself using one of your links, but I can’t figure out how you find the decimal that represents your score. Any input you have would be greatly appreciated! Thanks so much!

Response edited:

http://archive.csustan.edu/advstd/mat.html would give a raw score of 74 out of 100.

But since the mean is 400, the standard deviation is 25, and your score was 452, that would mean

Cumulative distribution function: 0.98123723

From http://danielsoper.com/statcalc3/calc.aspx?id=53

So you did better than all but 2% of the test takers, assuming that the scores are distributed in a normal distribution, with mean 400 and standard deviation of 25 points.

Sorry for the mix-up.

Hi John,

I have found this site to be very helpful in the weeks leading up to my MAT exam. In my experience, the official MAT practice exams were the closest approximation of the real thing. I scored 83, 82 and 81 raw on the three exams offered. I just got a preliminary score of 470 on June 5th. The raw to scale score conversion on the CSU Stanislaus page seemed to be spot on in my case. Thanks again for creating this informative blog.

Dave

Glad to hear you did well. Best of luck in your future endeavors.

John

Sorry to pop your bubble on the opening post, but the High-IQ societies do not accept the MAT.

You can put it in your application, but all that does it signal that you don’t meet their standards.

Yeah, okay, except that my MAT score got me into two of the most elite high-IQ societies.

Thanks for playing. Now go back to sleep.

Your post is very helpful. Thanks for sharing.

I am taking the MAT next Thursday, I am a little worried about it. The minimum requirement for my grad program is to get in the 50th percentile or higher, I am trying to figure out how many questions I need to get right in order to achieve that. Also, what raw score results in a 50th percentile. Your help would be greatly appreciated.

50th percentile w/b a score of 400 out of 600. Mean is 400, standard deviation is 50 pts, based on a normal distribution.

Did you check out this link from my post? http://archive.csustan.edu/advstd/mat.html

41 questions correct out of 100 gives you a score of 400, which is 50th percentile.

Good luck.

Hello again. I have links on the first post to my thread that answer your Qs.

http://archive.csustan.edu/advstd/mat.html

41 questions correct out of 100 give you a score of 400 out of 600, which is the mean/average score. Standard deviation is set to 25 points.

Basically, if you answer every other question correctly, you should be okay.

Good luck.

Hey John!

Just took the MAT, scored about 4 questions shy of where I needed to be. My question is if the official score is higher, lower, or the same as the prelim score. Thanks so much.

Hey Christina.

First off, congrats on doing a good job on this exam. Be proud of how you did.

As to your question, my understanding is that the preliminary score rarely changes, that is, it’s your actual, final score.

If this turns out to not be the case, please post here and tell us how different the scores were. (I really don’t think your score will change.)

Cheers.

Thanks for the quick response! I will definitely let you know! One more question, what does a 394 equate to percentage wise, I know that isn’t high like the everyone else’s on here, but it was pretty darn difficult for me lol Thanks!

394 means you answered 37 questions correctly.

http://archive.csustan.edu/advstd/mat.html

Given that the MAT scores are distributed as a normal distribution with a mean of 400 and a standard deviation of 25, then a score of 394 is the 40.5th percentile.

http://www.danielsoper.com/statcalc3/calc.aspx?id=53

Given a score of 394, this is under the mean of 400, which would have been the 50th percentile.