Page 1 of 1

US News rules, NRC drools!

Posted: Sat Jun 16, 2012 11:12 pm
by HappyQuark
While compiling a bit of data on graduate student earnings, I found an interesting trend between graduate school rankings and the number of applications universities receive. Specifically, there is a surprisingly good correlation between US News rank and applicants and, simultaneously, effectively no correlation between NRC S-Rank and applicants. It seems that either the vast majority of prospective students rely heavily on US News rankings or, perhaps, the same perceived strength as identified by professional physicists is shared by prospective students. Based on past and current conversations here, I tend to suspect the former rather than the later. I've posted the raw data here: http://www.physicsgrad.com/us-news-and- ... s-using-it

Image

Re: US News rules, NRC drools!

Posted: Sun Jun 17, 2012 5:08 pm
by bfollinprm
What are the vertical error bars?

Re: US News rules, NRC drools!

Posted: Sun Jun 17, 2012 6:13 pm
by CarlBrannen
The highly ranked schools tend to be larger. Maybe it would be more informative to have a chart that took that into account. Say applicants divided by number of PhDs granted, or number of grad students.

Re: US News rules, NRC drools!

Posted: Sun Jun 17, 2012 7:18 pm
by HappyQuark
bfollinprm wrote:What are the vertical error bars?
The US News ranks assigns universities a single value while the NRC gives schools a range score. The vertical error bars are just the range (min to max) NRC rank of the school. The curve fit for the NRC plot is doing its fit to the average (middle of the range) for each data point.

Re: US News rules, NRC drools!

Posted: Mon Jun 18, 2012 10:49 pm
by hb
HappyQuark wrote:
The US News ranks assigns universities a single value while the NRC gives schools a range score. The vertical error bars are just the range (min to max) NRC rank of the school. The curve fit for the NRC plot is doing its fit to the average (middle of the range) for each data point.
You have the NRC rank plotted along the horizontal axis, so the error bars should be horizontal. Unless they represent a range of the number of applications each school receives.

Re: US News rules, NRC drools!

Posted: Mon Jun 18, 2012 11:12 pm
by HappyQuark
hb wrote:
HappyQuark wrote:
The US News ranks assigns universities a single value while the NRC gives schools a range score. The vertical error bars are just the range (min to max) NRC rank of the school. The curve fit for the NRC plot is doing its fit to the average (middle of the range) for each data point.
You have the NRC rank plotted along the horizontal axis, so the error bars should be horizontal. Unless they represent a range of the number of applications each school receives.
Lame! You are right, I didn't realize I had done that. I fixed it and thanks for catching that.

Re: US News rules, NRC drools!

Posted: Tue Jun 19, 2012 12:57 am
by bfollinprm
HappyQuark wrote:
hb wrote:
HappyQuark wrote:
The US News ranks assigns universities a single value while the NRC gives schools a range score. The vertical error bars are just the range (min to max) NRC rank of the school. The curve fit for the NRC plot is doing its fit to the average (middle of the range) for each data point.
You have the NRC rank plotted along the horizontal axis, so the error bars should be horizontal. Unless they represent a range of the number of applications each school receives.
Lame! You are right, I didn't realize I had done that. I fixed it and thanks for catching that.
There ya go....
to be fair you should probably fit by weighting by half-widths (or any width, since weights scale like that). The uncertainty in ranking at the lower end is part of the reason that NRC rankings are harder to use as a prospective grad student (where the choice is often between schools whose ranges overlap).* I highly doubt that changes any conclusions, though.

*Which, of course, means that rankings are a horrible way to make these decisions. But, we crave numbers, and refuse to choice based on rational things like happiness and fit, and prefer to make our decisions on what others tell us is best, especially if they make a list for us!

Re: US News rules, NRC drools!

Posted: Tue Jun 19, 2012 4:23 pm
by misc123
Thanks for compiling this. I don't see UCSB listed though. Any reason, or just an oversight?

Re: US News rules, NRC drools!

Posted: Wed Jun 20, 2012 11:24 pm
by HappyQuark
misc123 wrote:Thanks for compiling this. I don't see UCSB listed though. Any reason, or just an oversight?
Based on the spreadsheet I put together, which was done a couple of months ago, the application numbers were not available on the gradschoolshopper page of UCSB. They seem to be available now so either they added them recently or I just somehow missed them the first time. I'll try to see if I can't update the table now that I have the relevant data.

Thanks

Update: I've added UCSB to the plot and table

Re: US News rules, NRC drools!

Posted: Thu Jun 21, 2012 7:51 am
by SSM
I can't believe I'm asking this, but what software did you use to plot these? They look nice.

Re: US News rules, NRC drools!

Posted: Thu Jun 21, 2012 6:51 pm
by HappyQuark
SSM wrote:I can't believe I'm asking this, but what software did you use to plot these? They look nice.
Quite amazingly, it's just Apples iWork Numbers. Typically I'll immediately jump to something like Mathematica or Matplotlib for plotting but in this instance I had everything in Numbers and I got lazy and left it there. For a spreadsheet program its a bit limited but the visuals are pretty impressive.

Screenshot Evidence