those graphs of schools, GREs, and GPAs

[One on One]

I love your site, thx! Also would be a good idea to maybe in the future track interests because admissions are surely random, but we can see which programs care more about interests to maybe uncover some of that randomness.

[untitled]

Originally Posted by One on One

I love your site, thx! Also would be a good idea to maybe in the future track interests because admissions are surely random, but we can see which programs care more about interests to maybe uncover some of that randomness.

Great idea. In fact, there sort of already is such a thing in the analyses page. Let me know if you have any other ideas, or if you wanted me to look at it in a different way.

[creepers]

@untitled: Quite an amazing site you've set up - brings out the necessary info from the crowds of threads. Thanks - it's really useful!

It'd be interesting if you can capture data on %age of international students applying and getting admits. It can also help answer the long-debated question on whether there's a skew towards int. students to bring in diversity etc. (Personally, I'll be happy if it is so )

[untitled]

Originally Posted by creepers

It'd be interesting if you can capture data on %age of international students applying and getting admits. It can also help answer the long-debated question on whether there's a skew towards int. students to bring in diversity etc. (Personally, I'll be happy if it is so )

Great idea. I've updated the csv file so it now includes a dummy for international student (not including Canada, which has its own dummy, so you can combine the two if you like, but including the UK and Europe, which I could take out, if that seems like a good idea).

Before checking, I would think that for our data, being international will be a bonus. Although Testmagic originally had an international focus, and thus attracted a cross-section of international students, I think that the econ board in the past 3 years attracts posts from only top international students, and the average international student will be slightly stronger than the average non-international.

However, upon checking, I'm not finding anything. I'll try to set it up so that I can easily output results of my tests from R into a format that is good for TM (anyone already have somehting simple worked up?); in the meantime, you can check out the csv file yourself.

[untitled]

As an example, here is an xtable output of a logit model for Harvard - does tm allow for tables? What would be a good way to present this information?

<!-- html table generated in R 2.8.1 by xtable 1.5-5 package -->
<!-- Tue Aug 04 13:10:06 2009 -->
<TABLE border=1>
<TR> <TH>  </TH> <TH> Estimate </TH> <TH> Std. Error </TH> <TH> z value </TH> <TH> Princeton Review(&gt |z|) </TH>  </TR>
  <TR> <TD align="right"> (Intercept) </TD> <TD align="right"> -38.3292 </TD> <TD align="right"> 30.7356 </TD> <TD align="right"> -1.25 </TD> <TD align="right"> 0.2124 </TD> </TR>
  <TR> <TD align="right"> yr09 </TD> <TD align="right"> 0.2488 </TD> <TD align="right"> 0.8375 </TD> <TD align="right"> 0.30 </TD> <TD align="right"> 0.7664 </TD> </TR>
  <TR> <TD align="right"> yr08 </TD> <TD align="right"> 1.1920 </TD> <TD align="right"> 0.8360 </TD> <TD align="right"> 1.43 </TD> <TD align="right"> 0.1539 </TD> </TR>
  <TR> <TD align="right"> gre </TD> <TD align="right"> 0.0456 </TD> <TD align="right"> 0.0386 </TD> <TD align="right"> 1.18 </TD> <TD align="right"> 0.2369 </TD> </TR>
  <TR> <TD align="right"> realanal </TD> <TD align="right"> 0.4275 </TD> <TD align="right"> 0.6283 </TD> <TD align="right"> 0.68 </TD> <TD align="right"> 0.4963 </TD> </TR>
  <TR> <TD align="right"> intl </TD> <TD align="right"> -0.5066 </TD> <TD align="right"> 0.7932 </TD> <TD align="right"> -0.64 </TD> <TD align="right"> 0.5230 </TD> </TR>
  <TR> <TD align="right"> canada </TD> <TD align="right"> -13.7424 </TD> <TD align="right"> 1455.3977 </TD> <TD align="right"> -0.01 </TD> <TD align="right"> 0.9925 </TD> </TR>
  <TR> <TD align="right"> ivy </TD> <TD align="right"> 1.9650 </TD> <TD align="right"> 1.1409 </TD> <TD align="right"> 1.72 </TD> <TD align="right"> 0.0850 </TD> </TR>
  <TR> <TD align="right"> public </TD> <TD align="right"> -0.3450 </TD> <TD align="right"> 0.7514 </TD> <TD align="right"> -0.46 </TD> <TD align="right"> 0.6461 </TD> </TR>
</TABLE>

[creepers]

I still am unable to locate the tables with 'international' as a dummy and the hyperlink links to the 'difficulty' table. Am I missing something?

[untitled]

Oops, I linked to difficulty.html rather than difficulty.csv. The link is fixed now. Thank you very much.

[untitled]

I've taken a (the?) plunge, and added logistic regression results to each schools page (for example harvard or washington). I've just gone with two models, one with everything (dummies for years, international students, private school students, students with real analysis, etc), and one with only GREQ and GPA. I don't use fixed effects (just starting that chapter of the book in my self study). I'm always happy to hear any suggestions. The R is pretty simple, here, so if anyone wants to tell me how to implement fixed effects or make any easy changes, feel free.

[untitled]

Here sample R code I'm using for the first regression. Everything is pretty simple. Does anyone know how to change the number of decimal places xtable reports? And why doesn't xtable give any model statistics (I'm using

xtable(anova(XXmodelXX))

to get those, which works fine, but if I had just null and residual deviance, that might be better)?

library(xtable)
data<-read.table("c:/scripts/difficulty.csv", header=TRUE, sep=",")

school<-data[,'.$Rschool.']
test<-school>=0
names<-data[,1][which(test)]
year<-data[,2][which(test)]
gre<-data[,3][which(test)]
gpa<-data[,4][which(test)]
gre<-(gre-mean(gre))/sd(gre)
gpa<-(gpa-mean(gpa))/sd(gpa)
realanal<-data[,5][which(test)]
intl<-data[,6][which(test)]
china<-data[,7][which(test)]
canada<-data[,8][which(test)]
ivy<-data[,9][which(test)]
public<-data[,10][which(test)]

yr07<-ifelse(year==2007,1,0)
yr08<-ifelse(year==2008,1,0)
yr09<-ifelse(year==2009,1,0)

school<-school[which(test)]

out<-summary(glm(school ~ yr09 + yr08 + gre + gpa + realanal + intl + china + canada + ivy + public, binomial(link = "logit")))
print(xtable(out), type="html")

[untitled]

Also, I have a ton of questions. First, is it obvious from what I've said and what I've done that it is totally unreasonable to use this for personal predicative purposes? I'm not sure what it can be used for, but I'm sure it isn't that.

Second, should I try to get null and residual deviance alone – it wouldn’t be hard, rather than show complete xtable(anova(.)) results? Or do people like looking at anova?

Third, how is the formatting? Should I put the regression output next to the chart, and move the most recently selected profile? Should regression results get their own file (I do plan on putting regression results for all schools into a single file, once I figure out if that makes sense, and how to organize it (heatmap?!?!))?

Does anyone love design web pages and want to design a better way to do it (I can do all the javascript and stuff, and I’m hesitant to use css in a webpage generated by perl, but, I’d love suggestions)?

And, as always, are there any obvious mistakes, or even any interesting trends that appear? Does anyone see anything interesting having to do with the dummies?