spacepenguin wrote:r6_philly wrote:FiveSermon wrote:
Always worse than using no data? Then why does anyone use lawschoolnumbers at all?
Is that really how you intend to address my objection? "My hunch is better because lawschoolnumbers is worse than using no data since it is flawed because it is a self selecting sample. It may indeed be accurate and be the best predictor we have but compared to my unflawed hunches and general musing it is obviously inferior hurr durr"
If you missed what I wrote before - you use it for dates, scholarship and such. You can even use it as a flawed individual predictor sure. But what you can't use it is to make assumption on the WHOLE population. The sample that is LSN is only representative of the population that is LSN. You do not use it as a sample of the whole LSAC population. The purpose of LSN would be what they call "for entertainment purposes only/not scientific" (even read the small prints on web polls?).
My hunch is based off of real published numbers - I read the published median/75s and they were lower than B/P and some below. It's a hunch, and I am selling it as such. You also have a hunch, but it is technically "baseless" because you are not considering the whole applicant population (by only looking at LSN which doesn't represent all applicants at all).
So your statements should be qualified as: For people who self-selected into LSN, it seems that CCN is in line with the other people's median ...
So just to prove a point, here is the published medians in question (for all applicants):
Columbia 3.72 / 172................(0, 0)
NYU 3.71 / 172......................(-.01, +1)
Chicago 3.78 / 171.................(+.02, 0)
Stanford 3.88 / 170................(0, 0)
Penn 3.85 / 170.....................(+.03, 0)
UVA 3.85 / 170......................(0, 0)
Duke 3.80 / 170.....................(+.04, +1)
See how LSN isn't worth crap to predict the whole pool?
I am serious, you must not have taken stats. I think the 1st or 2nd chapter usually talks about not using non-random samples and how bad data can mess things up.
I'm confused, are you comparing the law schools' published medians versus LSNs medians? If so, the law school medians would be reflective of those who are attending while LSN's would be of those merely accepted and not those who matriculated. Also, while the data is different, have you actually demonstrated a difference of statistical significance?
Non-random samples are bad, I agree, but the normalization and form of self-selection through LSN should lead to a positive bias rather than a negative one IMO.
Well according to R6, the entirety of LSN is less predictive as a whole than his acute intellectual reasoning seasoned through precious minutes of reading an article which happened to mention something about something...about something...