MS stats: UWM free vs Duke paid

[Heathcliff1850]

I've applied to MS statistics programs, and I have acceptances at the University of Wisconsin at Madison (UWM) and Duke.  UWM will let me go for free, whereas at Duke I must pay full price ($60k/year for two years).  Which choice makes more sense?  I'm leaning UWM, but I'm wondering what other people think.

I'd like to go on to a PhD.  From scanning lists of alumni, I note that both UWM and Duke historically send a lot of MS students to PhD programs, which is a point for both programs; UWM PhD in particular seems to take a lot of people from its own MS, more so than Duke, but I also worry that, if I want to leave my MS school, Duke will give better opportunities.  UWM's MS doesn't have a thesis option and Duke's does, so point Duke, though both seem to have pretty flexible curricula, so maybe I could do a thesis at UWM anyway.  Tuition at Duke is $60k per year and the program is two years, and $120k of debt is rather a lot, especially given that I already have debt from a previous degree; big point for a fully funded degree at UWM.  UWM has promised a TAship with a $16k stipend, but Duke says that TAships are generally available to MS students who want them, so maybe that's a wash.  I love Bayes' Rule, and if I could jump into a tub full of Bayes and roll around in it for the rest of my life, I would be happy; while Duke is obviously superior in that respect, UWM has people doing cool Bayesian stuff in biomed, and anyway, at the MS level, maybe the most important thing to prepare me to roll in Bayes at the PhD level is to be really, really good at probability theory, which I could learn at UWM.  Finally, Duke outranks UWM by 0.1 points in the US news rankings (12th vs 16th), which seems like not enough to worry about.

In sum, it seems to me that Duke is a little better for what I want, but not $120k better, and that I should go UWM.  Gradcafe, what do you think?  (Sorry for the slightly manic post.)

Edited April 14, 2018 by Heathcliff1850

[jswizzle48]

I think it's great that you are voicing your options. I think the best bet for you is to go to UWM. It's an incredible school, with a fantastic program. Considering you have a lot of debt from your previous degree, I would agree with you that attending Duke would not be wise. 

 

Congratulations on being admitted to such great schools

[StatHopeful]

I would advise you to take a funded MS offer at a 30-50 rank institution over paying 120k for a 'top-tier' MS any day. You're talking about UWM, an elite program in its own right. 

Assuming you only have 20k in debt now and all of your debt was on a subsidized Stafford (impossible but way better situation than anything you could end up in), at a 4% rate you'd be making essentially a mortgage payment for ten years.

It's also important to note that the mountain of debt you're going to have coming out of Duke is going to pay a huge role in consideration of PhD's on the way out. Being as financially free as possible will allow you to select the department where you'll be happy and do your best work.

There's no way I'd recommend going to Duke in this scenario.

 

Edited April 14, 2018 by StatHopeful

[bayessays]

No brainer, go to Madison. 

[Bayesian1701]

UWM all the way.  I am a Bayesian and I would make that decision if I was in your shoes.  Duke is better for Bayesian but not $165k after interest better.   Plus UWM is high enough ranked to probably help you get in a good PhD program in two years.

[Heathcliff1850]

Let's play a game!

 

Let M denote the event that I should go to UWM, and let D denote the event that there is an 11-0 poll majority telling me that I should go to UWM.

I said that I was "leaning UWM".

  p(M) = 60%

If UWM were the better choice, what are the odds that exactly 11 people would tell me so, and no one would tell me otherwise?  I guess I could see the vote being 10-1 or 9-2.

  p(D|M) = 70%

If Duke were the better choice, what are the odds that no one would tell me so, and exactly 11 people would tell me otherwise?  Seems like a stretch...

  p(D|M') = 2%

Applying our favorite formula, we have:

  p(M|D)

  = p(D|M) * p(M) / p(D)

  = p(D|M) * p(M) / (p(D|M) * p(M) + p(D|M') * p(M'))

  = 70% * 60% / (70% * 60% + 2% * 40%)

  = 42% / (42% + 0.8%)

  > 42 / 43

  > 39 / 40

  = 97.5%

 

 

So yeah, I think I'll go to Madison.  Thanks to everyone for the input.

Edited April 15, 2018 by Heathcliff1850