You've misinterpreted my statements left and right.
Originally posted by skidmark
Jed,
Based on your reasoning the average gpa for a medical student after 2 years should be around 3.6. I don't know about others, but this is not the case for my school. Probably no more than 10 people in my class have ever made more than a handful of Cs as an undergrad. Yet, this past semester each course had over a quarter of the students (25) make C or worse. Four people even failed out. How many entering medical students do you know that failed college semesters?
This is a basic statistics issue. If you are going to try to make a predictive model, or equation, showing a relationship between two groups of data you have to account for the mean and standard deviation of each group. Med school means are not usually equivalent to undergrad means, so, ipso facto, any model will not predict such a result.
Originally posted by skidmark
Yes, GPA and MCAT are used for screening applicants but they have little correlation to your grades in medical school. (USMLE scores are a different story, they actually do correlate w/ MCAT). What is the difference btwn a 3.5 in biochem from one school (lets say Harvard) and a 3.8 in microbiology from another (say LSU)? Answer: Not possible to answer. Your reasoning suggests that the guy from LSU will do better in med school the majority of the time. What is the formula?
As far as admissions, I think it's clear that a 4.0 from Podunk College isn't equivalent to a 4.0 (or probably even a 3.5) from an elite school, whether public or private. Selectivity of one's undergraduate school is usually factored into medical school admissions. This can be done by a model with different weights for different schools (1 to 5, or whatever) or it can be done (probably more commonly) by simply using judgement.
As far as predictive value of academic markers (for predicting med school performance), you have a point. Models used by researchers should incorporate the difference in undergrad schools. But I haven't seen one that does. All of them just use straight GPA, which, despite being imperfect, does add to the R^2 value (proportion of variance explained by a linear regression model). Bottom line, GPA, even unadjusted for institution, does add predictive value.
If you are really interested in this stuff, you should access the primary literature.