uranium wrote:knowing that 10,000 students in the world take the PGREs every year, how many do you think you'll need to gather a meaningful sample for statistical analysis? ...

61 participants per type of test are needed to sample the whole results space. There are only 61 types of scores ranging from 990 to 390 in steps of 10. A statistical analysis considering the 10,000 students is not necessary because ETS already did it for us

. We only want to gather the results.

By the way, we don't need to sample the whole results space to make the comparison between different types of exams. At least I would be more than satisfied if we compare the difficulty of achieving scores between 990 and 700. We don't even need all consecutive results. We could just interpolate some.

uranium wrote:Finally, whose interest is it in to have this analysis?.

It's mainly curiosity. It's a physicists forum, what do you expect? Not every pursuit of knowledge needs a direct application.