cee cee wrote:Uh, I'm no stats guru. But the way I interpreted it: say you apply in December and you have an 8% chance of being accepted to Chicago. Had you applied in November, you would have a (8% + (78.8% of 8%) ) = 14.3% chance of being accepted to Chicago. Applying in October would give you a (14.3% + (78.8% of 14.3%) ) = 25.6% chance of being accepted. Am I thinking about this the right way? This still seems super high, but with a bias towards large magnitudes, this just makes a stronger case that Yale does not advantage earlier applicants.
I'm not sure that's what he's saying. I think he's saying that in any given earlier month (Sept vs Oct, or Oct vs. Nov) at Chicago, for example, you have a 78.8% better chance of being admitted. I don't think it gets compounded the way you're describing.
But either way, there are tons of problems with this analysis (some of which he acknowledges, some of which he doesn't).
(1) It comes from self-reported LSN data. We like to believe that data is at least somewhat representative as a sample, but we have no way of knowing. And all it would take to blow up the model is something like 10-15 missing applicants with very different numbers.
(2) The Sept. 30/Oct. 1 problem that he talks about. Basically his model draws a line at change in month, but change in month could be just one day. So to say there's a huge boost for applying one day earlier is ridiculous.
(3) What he doesn't talk about is that there's a selection bias problem. While we don't know for sure if this is true, it's possible (I would say likely) that early applicants are stronger applicants. Among other things, they are more likely to be applicants who are not retaking the LSAT, which means they are applicants who don't need to retake the LSAT. Also, for lack of a better term, they are applicants who have their shit together. They got their letters lined up, personal statements written, etc. all nice and early. Obviously that doesn't guarantee that they are better applicants, but it certainly raises the likelihood that some of them are.
TL;DR - There may be a boost for applying early but that regression analysis, to me, doesn't do much to support the hypothesis.