<p>
</p>
<p>I think this is true of only one of the rankings (the “rank-ordered logit of preference orderings”, or some such title). The main ranking is cooked from a hybrid of the matriculation decisions by students admitted to two or more schools, and (as a proxy for the matriculation decision in the case of Early Decision applicants), the applicants’ ranking of colleges on the questionnaire. </p>
<p>None of this affects your point, which is completely correct, that citing the NY times chart as “data” or even “based on data” in any direct sense, is nonsense.</p>
<p>
</p>
<p>It <em>is</em> just a computer model. How else would you describe it?</p>
<p>
</p>
<p>Only one out of the four (the statistician) appears to have the background needed for developing this type of ranking algorithm and interpreting the results. The others brought the funding and the data.</p>
<p>
</p>
<p>One reason they needed a high-powered statistician was to squeeze what they could from a small sample. They had 3240 data points to measure around 150 parameters. The number of data points is maybe half of that in the ranking based on the science and engineering applicants, the one that showed Yale beating MIT (game over!) just as it did in the overall ranking. Below the top the rankings are totally unstable due to the lack of data (sparsity of the matrix of cross-admit results). At the top, they are sensitive to the sample in other ways, as seen in the Yale and Caltech rankings.</p>
<p>
</p>
<p>Harvard has said that MIT is their toughest competitor and that they win by 3:1 margin against the other rival schools. MIT beats all others in cross-admit battles, with Caltech and Stanford (maybe also Princeton, I don’t remember) being the main rivals other than Harvard. If you don’t believe what the ex-admissions worker told you in this thread you can glean the above from sources locatable with a search engine. </p>
<p>
</p>
<p>The survey of all admitted MIT students, not just the enrolled freshmen, had a 92 percent response rate a few years ago. Apparently web searching fills most of the remaining gaps.</p>
<p>^^Yeah, on the MIT admissions reply card, they ask what school you are going to if you turn down MIT.</p>
<p>I looked at the NYT table mia linked, and it doesn’t claim to be a cross-admit battle. I would guess that almost all humanities majors would prefer Yale (with the exception of economics.) In the NYT simulation, these people would prefer Yale over MIT. However, most of these people would not even apply to MIT. For this reason, the revealed preference does not correlate to cross-admit battles.</p>
<p>The NYT chart calls it “data” and uses a factual tone (“66 percent of students who were accepted at both Duke and Penn … went to Penn”). Although they mention that the “data” are estimated from a statistical model based on a survey, that is totally misleading.</p>
<p>NYT readers know that a national political poll involves some statistical models that might crunch the numbers in a more complicated way than just adding up the statewide totals, so if the survey says 55 percent support Obama, that is not necessarily the fraction observed in the poll. However, everybody also understands that the sole purpose of the statistics is to accurately estimate that number.</p>
<p>What is misleading here is for NYT to not make clear that the numbers they reported were not something the Revealed Preference study attempted to measure or estimate whatsoever. The study claims no particular predictive power for cross-admit battles.</p>
<p>The NYT also did not make clear that the model predicts much different, and often strange, results for the real-life type of cross admit battles, where a student is admitted to <em>more than two</em> schools and has to pick one. The table in NYT is about a hypothetical decision for an applicant forced to choose one school or the other. This is different from the decisions of those applicants who happen to choose one or the other, but also had other options. The latter is the situation that NYT purports to be talking about, but it is not what “cross-admit probabilities” refers to in the study.</p>
<p>The New York Times did spawn a lot of nonsense with this chart.</p>
<p>Harvard > Princeton > Yale > Stanford > MIT</p>
<p>It might be HYPSM, but it should really by HPYSM.
I guess saying “hip sum” is more catchy than “hpism”</p>
<p>^And do you have enough data to back your random claim?</p>
<p>Different schools have different strong points, so if you are talking about humanities, MIT would be last, but if you’re talking about engineering, I think you would have to move those schools around.</p>