Percentile Calculator

[ggmk]

Advertisement - Members don't see this ad

There's been many posts about figuring out the percentile of someone's score given the mean and SD.

Hope this could settle the confusion once and for all.

Here's a calculator for percentile:
http://bayes.bgsu.edu/nsf_web/jscript/normal_cdf/normal_icdf.htm

On the left, put in the mean and SD.

On the right, under PROBABIILITY, input the percentile in decimle points. Like .90 for 90th%tile, .30 for 30th%tile

then click COMPUTE X, it will tell you the score that corresponds to the percentile you input.

For exmple, COMLEX mean 500, SD 79, using probabilty .90, then x = 601.
So a 601 corresponds to a 90th%tile.

 

[SquidDoc]

Interesting...I wonder how my 2 digit score is 6 points lower than my percentile per the calculator. I guess the 2 digit score is more of a raw score than a percentile estimate.

 

[Idiopathic]

The two digit COMLEX is going to be way lower than the percentile. The two digit is like based on a mean of 80 with a 3 or 4 point SD, so my 92 was actually way higher than that percentile wise.

Unfortunately, they report the two digit to residency programs, so my XXX/99 USMLE looks a lot more impressive than my XXX/92 COMLEX, even though the latter is a better relative score.

 

[flipcyde]

dont' post often but been following sdn for many years. thanks for the link, just curious as to what other people think of this website. how accurate do you think this analysis is, a score of 601 or greater puts you in the 90th percentile? these posts of >650 amaze me, putting you at the very top of the pyramid with an impressive 97% percentile. a score of 700+ according to this website puts you at the 99.5th percentile, if that is indeed true, rock on you studs! i did well but no where close to that incredible score.

 

[Mallory Weiss]

dont' post often but been following sdn for many years. thanks for the link, just curious as to what other people think of this website. how accurate do you think this analysis is, a score of 601 or greater puts you in the 90th percentile? these posts of >650 amaze me, putting you at the very top of the pyramid with an impressive 97% percentile. a score of 700+ according to this website puts you at the 99.5th percentile, if that is indeed true, rock on you studs! i did well but no where close to that incredible score.

Don't worry, as long as you did well I'm sure you will not have a problem. No matter how well you do anything, there will always be someone who does it better than you. That is just a fact of life, so don't feel bad about it. When you think about it this way, it sounds better: A score of 601 puts you in the TOP TEN PERCENT of scores of some of the smartest people in the nation. A score of 500 puts you in the top half. And as long as you pass, you can get some residency somewhere.

 

[osli]

That calculator only works for a normal distribution, and I'm sure USMLE and COMLEX scores don't follow a normal distribution. Though it might be good enough for a quick and dirty guesstimate of your percentile.

 

[Pinkertinkle]

Don't worry, as long as you did well I'm sure you will not have a problem. No matter how well you do anything, there will always be someone who does it better than you. That is just a fact of life, so don't feel bad about it. When you think about it this way, it sounds better: A score of 601 puts you in the TOP TEN PERCENT of scores of some of the smartest people in the nation. A score of 500 puts you in the top half. And as long as you pass, you can get some residency somewhere.

 

[werd]

That calculator only works for a normal distribution, and I'm sure USMLE and COMLEX scores don't follow a normal distribution. Though it might be good enough for a quick and dirty guesstimate of your percentile.

indeed. this has been discussed before; while there's no conclusion to the debate, the general belief is that the usmle (i don't know as much about comlex) does not have a "normal distribution of scores." therefore, one's percentile cannot be acutately calculated based on avg + st. dev.

consider 1) the usmle abolished percentile reporting citing that it's misleading. they say that a 220 one year is the same level of performance as a 220 the next year, even if the mean or st. dev has changed. the number of people failing also changes from year to year, suggesting against normalization. this implies that scores are not normalized to large groups of test takers, but generated internally on an individual basis of performance.
also, 2) logically... about 10% of people fail step 1. on my test year the avg was 217 and pass was 182. assuming a normal curve, 10% scored at least 35pts below average, so 10% must also score at least 35pts above average. this implies that 10% of people scored 217+35= 252 or higher. maybe that's true on the sdn forum, but it just doesn't ring true for what i know about scores.
finally, if the distribution were normal, the usmle wouldn't be witholding any data from us by abolishing percentile reporting. most with a college (high school?) degree can calculate a percentile from a mean and standard deviation.

 

[Law2Doc]

ROFL.
Yes, when you (MW) resurrect a two year old thread, please say you are resurrecting a two year old thread. That way folks don't bother replying to posters who probably aren't reading the USMLE Step 1 board anymore (since they are probably licensed by now.

 

[sprinkibrio]

I thought this thread would be helpful for those of us who just got our scores today and are wondering about approximate percentiles...

 

[SRK Jerk]

I don't think the USMLE follows a normal distribution. As was mentioned, if it did, why not just release percentiles? Anyone can make this calculation with the mean and SD. Can't be that simple.

Also, if 250 was really the 90th percentile, I doubt it would be considered as great score as it is.

 

[sprinkibrio]

Percentile doesn't mean a lot, but I don't think it's because the tests are skewed. A good standardized test forms a normal curve. I think percentiles are not always accurate because not all administrations have the same range or mean if the actual score is supposed to be the same whichever test form you took. For example, if you took the test at a random time like August with IMGs and people who failed the first time, you would have a higher percentile with a 140 than if you took the test in June with most US students. Also, if you took the test in December, the range would be much smaller since fewer people took the test. If you got a 140 you might be in the top 5% of that administration.

I think percentiles mean the most when you're taking the test in June with 90% of the other US students who you'll be competing with for residency.

 

[deeplyinitnow]

Percentile doesn't mean a lot, but I don't think it's because the tests are skewed. A good standardized test forms a normal curve. I think percentiles are not always accurate because not all administrations have the same range or mean if the actual score is supposed to be the same whichever test form you took. For example, if you took the test at a random time like August with IMGs and people who failed the first time, you would have a higher percentile with a 140 than if you took the test in June with most US students. Also, if you took the test in December, the range would be much smaller since fewer people took the test. If you got a 140 you might be in the top 5% of that administration.

I think percentiles mean the most when you're taking the test in June with 90% of the other US students who you'll be competing with for residency.

I took the exam in june and got a 237 (99th percentile). I took it on a sunday... could have played a role

 

[Glaxo]

I took the exam in june and got a 237 (99th percentile). I took it on a sunday... could have played a role

congrats on the good score but the 99 they give on the score report is just another scoring metric they use not a percentile