<p>This is scary. Using Prestige's updated data (post 115), and calculating simple and multiple regressions. All the factors (USNews ranking, WSJ, RP, and % NMS) are highly significantly correlated p<0.0005. The lowest correlation was for RP rank vs NMS rank (0.62). If one wanted to predict WSJ rank (conceding all the problems with the WSJ), then each of the other factors would produce a highly significant correlation. The highest correlation was with the USNews average rank. In fact, if one were to predict the WSJ rank using all three other factors, then, with USNews in the model, RP and NMS are no longer significant. So whatever the USNews does includes the predictive value of RP and NMS, if the goal is predicting who will end up in these professional schools.</p>
<p>Repeated deleting Caltech, since it is an outlier for WSJ based on its ranking on the other scales- same result. The much, and deservedly, maligned USNews contains all the predictive value held in the combination of USNews, RP and NMS.</p>
<p>So, if your main interest is in where your classmates are likely to go to prof school, then you can use the USNews, for up to date information that correlates very well with the WSJ, or figure that the WSJ does not change much, and use it.</p>
<p>RP and NMS are highly related (after all the schools with high NMS are like that because these sought after students choose to go there), and do not contribute to knowing where people go to prof school if you know USNews or WSJ. However, either might tell you which colleges are popular among top students.</p>
<p>Repeated using only the top 15 from Prestige's list. Here the multiple regression for predicting WSJ just misses significance (p=.06). So the high predictive value of these models is for the large differences between those at the top and the bottom of the full set of 27. Considering only the top universities, none of the predictors reaches significance for predicting WSJ, although they come close, and USNews is again the best.</p>
<p>Since all the factors are highly correlated, changing the weights does not have much effect on the overall ranking, but of course might be useful for a troll who wants to proclaim his/her favorite college is "better" than some other. </p>
<p>None of this tells which is the "best college", there is no such animal.</p>