(Pearson) Correlation is the wrong method. Outcome is binary. Use proper model and it works fine. http://taxprof.typepad.com/taxprof_blog/2013/09/law-school-gpa-.html …
-
-
Replying to @KirkegaardEmil @robvfour
Effect sizes matter; looking at the regression table (and descriptions) confirm result looks weak
1 reply 0 retweets 0 likes -
Plot shows the predictions to be quite useful. Validity mostly obscured by the low bar for passing the test.
1 reply 0 retweets 0 likes -
Replying to @KirkegaardEmil @robvfour
Plot shows the opposite to me! You have to jump really far across the range of LSAT scores for big changes in probability
1 reply 0 retweets 0 likes -
Take the -1 z GPA (least restricted). There's 7 ticks. Compare 3rd vs. 5th, for a +85% increase in pass rate.pic.twitter.com/ch4xsXSG2E
1 reply 0 retweets 0 likes -
Replying to @KirkegaardEmil @robvfour
You know that using % increase rather than % point increase is highly misleading
1 reply 0 retweets 0 likes -
Not misleading in this case. OR only superior to RR when base rate is very high. I chose the least problematic curve for this reason.
1 reply 0 retweets 0 likes -
Replying to @KirkegaardEmil @robvfour
The curve isn't the issue! Your misleading portrayal of the effect size is.
1 reply 0 retweets 0 likes -
Was not misleading. Already explained this to you.
1 reply 0 retweets 0 likes -
Replying to @KirkegaardEmil @robvfour
You misinterpreted my complaint as an OR vs RR issue
1 reply 0 retweets 0 likes
Next time, use blog posts etc. to discuss stuff. Twitter is terrible. No idea what your point is. Effect size is clearly substantial.
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.