Are you familiar with US News’ top 25 national universities?

[Sam_Lee]

<p>

In order to observe the dynamics of the top 25 schools, I subdivided the USNews’ rankings into three periods:</p>

<p>First periods fell between 1983 and 1989(Y1980s),
Second periods between 1990 and 1999(Y1990s), and
Third periods between 2000 and 2009(Y2000s);
And then averaged them to obtain Rank1980s, Rank1990s, and Rank2000s (see post #1’s tables for details). </p>

<p>Schools that ranked “top 10 for at least two of the three periods” were defined as institutions which ranked “top 10 for at least two of the three rankings: Rank1980s, Rank1990s, and Rank2000s”.</p>

<p>Rice ranked 12, 16, and 17 for Rank1980s, Rank1990s, and Rank2000s, respectively; it belongs to tier 4, a highly competitive level.

</p>

<p>Why is subdividing this way is better than any other ways in terms of “observing the dynamics of the top 25 schools”? </p>

<p>Also, it seems like all the more time-consuming work you were doing is actually inferior to simply averaging all the years’ rankings for each school and then assigning tiers.</p>

[whartongrad08]

<p>this has to be one of the dumbest threads I’ve ever read on CC…I want my 10 minutes back.</p>

[modelingLiao]

<p>SamLee,</p>

<p>For 1983, 1985, 1988, and 1989, Berkeley ranked 5, 7, 5, and 24; Cornell ranked 8, nr, 11, and 14; and Northwestern ranked nr, nr, 16, and 17. For the first three USNews rankings (1983, 1985, and 1988), Berkeley was among the top 5 but dropped to 24 one year later (in 1989); Cornell initially ranked at 8, then went unranked, and then reappeared at 11 (in 1988); and Northwestern was mysteriously left out for 1983 and 1985, but came back strongly ever since. It is not easy to perform trend analysis for data of this level of inconsistency.</p>

<p>My approach: USNews’ rankings were grouped into three periods, defined as Y1980s, Y1990s, and Y2000s, each encompassing a decadal interval (except for Y1980s) in order to model the movements and project the trends of schools’ rankings, which is difficult to draw meaningful conclusions based solely on year-by-year data. Tables on post #1 provided me with the decadal rankings breakdowns for the past 30 years, which is more informative for overall ranking assessment in comparison with the use of overall averaged ranking. </p>

<p>Let me use those four schools (Berkeley, Cornell, Northwestern, and UVA) you mentioned in your previous two posts as an example. For Rank1980s, Rank1990s, and Rank2000s, Berkeley ranked 10, 19, and 21; Cornell ranked 11, 11, and 14; Northwestern ranked 19, 14, and 12; and UVA ranked 18, 20, and 23. Again, the above-mentioned tier-ranking results were derived based on the following five criteria:
Tier 1 (1-5). Universities ranked among top 5 for at least twice of the three periods
Tier 2 (6-10). Universities ranked among top 10 for at least twice of the three periods
Tier 3 (11-13). Universities ranked among top 10 once for the three periods
Tier 4 (14-20). Universities ranked among top 20 for at least twice of the three periods
Tier 5 (21-26). Universities ranked among top 20 once for the three periods and/or Universities ranked among top25 for all three periods.</p>

<p>I put a lot of weight on top 10 and top 20 as the second and third cut criteria (the fist cut criterion was top 5). The former resulted in the induction of Berkeley (barely made it at 10 for Rank1980s) and rejections of Cornell (closely missed with 11 and 11 for Rank1980s and Rank1990s)/Northwestern. The latter caused the separation between UVA(narrowly escaped with 18 and 20 for Rank1980s and Rank1990s) and other elite schools listed on tier 5, e.g., Notre Dame, CMU, Michigan, Georgetown, UNC, Vanderbilt (6 in total). </p>

<p>One may redefine or redesign different criteria in the model and come up with different tiers results; but that’s another different story. The approach I presented on this thread can be used for trend analysis and ranking projection, which reflecting historical data (rankings).</p>