Correlations UW, UWSA, NBME vs Step 2CK
First time user of this forum. A lot of spread with the data here, so I went through all the data, collected it into a master file, and ran some correlations. See the attached images and data file.
I focused on UWorld first pass percent correct, UWSA, and NBME. For UWSA I recorded the data no matter when the assessment was taken. However, I only ran the correlation for UWSA scores <265, since it seems to max at that value and that would skew things. For NBME, I only used assessments taken near to the actual exam date, and generally did not include NBME form 7. Not super scientific here, folks. The correlation thus will most hold for you if you are taking a non-form 7 NBME close to your exam date. The line of identity is indicated by the dashed line (i.e., if the assessment perfectly predicted actual score).
A few findings:
-Small sample size with restriction of range effects and clearly a selection bias wrt who posts on this site.
-There's a lot of variability (see the R2 values).
-An individual test is helpful in ballparking, but do not assume that will be your score.
-For UWSA, it tended to overestimate scores above 240 but underestimate below that cutpoint.
-For NBME, it tended to underestimate scores across most of the range, and became unreliable above 260.
-For UWorld first pass %, I was surprised at the strength of correlation, but still tons of variability there! Hard to say how this data applies at the individual level, since many folks use UWorld at different times / in different ways. However, if you're getting 75%+ on your first pass you are in good shape for a score >240. If you're getting 65%+, most scored >220.
Hope this is helpful. Now I need to get back to studying...