« Two quick hits: how bad data analysis harms our discourse | Main | The plural of anecdote is not ... »


Feed You can follow this conversation by subscribing to the comment feed for this post.


How do you know that the retests are a less "accurate measurement of whatever the GMAT test is supposed to measure,” as opposed to the original sittings?

Sure, if the goal is the sort people, having different numbers of testings literally violates the very defintions and assumptions of “standardized” testing. Yes.

But if the goal is to sort by a particular construct, and to try to accurately measure that construct in order to meet that goal, how do you know that the restings are not increaingly valid measure. In theory, retestings oculd get rid of construct-irrelevent issues like lack of familiarity with the test format and stressed caused by the stakes of the test, caused by the unfamilar setting, etc. etc.

Yes, there is certainly something unfair about SOME students retesting more than others, but maybe the lack of validity is due to the one who do NOT retest, as opposed to the ones that do.


ceolaf: I am not against retests. In fact, statistics would argue that retests increase the sample size and provide a more accurate measurement. My comment is related to the various policies surrounding the retests and the reporting of the scores. You pointed out one key issue, which is that only low scorers are retesting today. It is worse than that: anyone who gets an unexpectedly high score will retain it and not retest. The new rule allows removal of low scores from the sample, which is yet another measurement issue.

myles gartland

I think you missed a factor here. Yes, the top 10-20 schools are chasing GMAT scores, but the trend is to waive (or make optional) the GMAT/GRE for many accredited programs. What you then get is the people who would do well still take it, and the ones who would not get a 600+ find other options for their application. Last grad event I went to at least 50% of schools have some kind of GMAT waiver. So, the average score (that GMAT reports) will go up based on self selection.


Ceolaf: From a statistical perspective the tests measure something, hopefully something useful, but with effort. The error results from choice of questions, day to day performance and a bit of luck. Repeatedly taking the test results in a GMAT from when teh error is high and positive.

There are good reasons for allowing a retest, as students may not be prepared the first time, or they may have a poor day. Poor preparation is something that shouldn’t be an excuse for someone doing grad school. However repeat tests should be limited to one or students should be only allowed to delete scores before they have seen the test results.


Myles: Thanks for the comment. I didn't know that GMATs are going out of favor. Perhaps they are making up the revenues by encouraging retests?

Robert Smith

As per my opinion, the GMAT score itself is not as important as what percentile that score represents. A few people score very high and very low, and the majority have scores in the middle. If more people score higher, then the mean score becomes the higher score and the bell curve then shifts to the right. (Similarly, if more people score lower, the curve would shift to the left.) This then shifts the percentile rank score.

The comments to this entry are closed.

Get new posts by email:
Kaiser Fung. Business analytics and data visualization expert. Author and Speaker.
Visit my website. Follow my Twitter. See my articles at Daily Beast, 538, HBR, Wired.

See my Youtube and Flickr.


  • only in Big Data
Numbers Rule Your World:
Amazon - Barnes&Noble

Amazon - Barnes&Noble

Junk Charts Blog

Link to junkcharts

Graphics design by Amanda Lee

Next Events

Jan: 10 NYPL Data Science Careers Talk, New York, NY

Past Events

Aug: 15 NYPL Analytics Resume Review Workshop, New York, NY

Apr: 2 Data Visualization Seminar, Pasadena, CA

Mar: 30 ASA DataFest, New York, NY

See more here

Principal Analytics Prep

Link to Principal Analytics Prep