Quant Preparation

[Canadianna]

I'm applying for Public Policy PhDs for Fall 2014. To begin, a disclaimer: I'm not sure if this is where I should be posting - it seems that the Government Affairs forum is almost exclusively Masters applicants, so this seemed more appropriate.

 

The programs that I'm applying for didn't specifically dictate anything beyond that I should have some stats and econ in my background, but now that I'm mostly done with applications and just working 9-5 for the time being, I was wondering whether I'd be well served to take a quant night class at the local university.

 

As for my current background, I took Calc 1000 way back in 2007 (I don't remember anything from it), and then I took Stats 2000 in 2010. I also did intro micro and macro economics. During my masters I took a research methods course that was for all intents and purposes a stats course (we used Stata and did basic analysis of data sets... finding the strength of different variables and such), plus I took public sector economics, international trade and environmental economics.

 

So my question is whether I would be well-served to take anther quantitative type course before I (god willing!) head off to start my PhD. I checked out the local university, and I could retake Calc 1000 as a night class, take Calc 1001 as an online course or probably do a small assortment of econ classes, either online or at night. No stats classes seem to be offered at convenient times, but I could always take one during the summer or online from a different university.

 

Financially, I can afford it but would rather not waste money on it if it wouldn't be helpful.

 

Any thoughts would be appreciated.

[CGMJ]

I can't speak to public policy, but if it were me I would take classes in statistics/econometrics. The more exposure the better, even if you aren't planning on being a methodologist. At the entry level at least, you need minimal calc and linear algebra, so if you're already familiar with calculus I think stats courses would be much more useful. 

[catchermiscount]

The further and further down the technical rabbit hole you go, the deeper and deeper the texts you read become.  And yet, the deeper and deeper the texts are, the more and more they tell you that no prerequisite is required other than "mathematical sophistication," which isn't taught in any formal class.  In another thread, RWGB---who is far more qualified to speak on these matters than I am---mentioned a good book on writing proofs.  Proof-writing is a wonderful skill to develop in that it forces you to think about how assumptions and conclusions (whether you had them in mind to begin with or not) link together.  The more sophisticated you are, the better you become at picking up skills as they are taught to you.  So, in the event that you decide to cut costs and work on stuff by yourself, I would suggest that you pick up a basic math book (Simon and Blume, e.g.) and a basic proof-writing book (RWBG's suggestion was good, or The Nuts and Bolts of Proofs).  The best case scenario is that this stuff becomes intuitive and fun---a play experience, really.

 

As I read it, this advice might seem a little more suited for theory than for statistics.  That's probably true in a very applied sense. 

[IRToni]

I would definitely say that the proof-writing advice is great. Having done a minor in math in college, I can definitely say that proof-writing has helped me also in my non-math studies. However, if you're planning on doing public policy and especially statistics, actually learning stats might be more useful for you (in terms of opportunity costs). While my econometrics class included a lot of proof-writing, the applied part (using a statistical software) was actually what I found extremely useful. Of course, knowing the assumptions that need to be fulfilled (true) is necessary, but assuming you haven't done much beyond linear regression/OLS in stats, this might be more useful than proof-writing as such. There are tons of online (free) courses available that could also teach you R, it would only depend on you being able to get yourself motivated, I think.

 

I wouldn't say it's necessary, though. Whatever you do know might only make the first couple of weeks/months of classes in your PhD program easier, but probably won't make or break it, as you seem to have had quite some exposure to quant stuff in the past. I'd also think of the opportunity cost, because, assuming you'll get in, you'll be spending a lot of time doing school-related stuff from September onwards, so focusing on some private project that you won't have time to then might be more cost-effective. I will say that I think learning R would be worth-while for me, and I am thinking of taking a course/teaching myself before the PhD.

[RWBG]

RWGB---who is far more qualified to speak on these matters than I am

 

Well this definitely isn't true.

I think if you've already applied (so there isn't a signalling component) and you're just trying to learn some things to help you engage with you Ph.D coursework, then proof-writing isn't too bad an option. More generally, while I wouldn't disagree with IRToni that applied statistics may be more helpful to your research than proof-writing, when assessing opportunity costs, you should keep in mind that it is very likely that your Ph.D program will spend time teaching you statistical modelling in a way that they may not spend time teaching you proof-writing, or, say, linear algebra. So, optimizing over the longer term, you may be best served by covering content that will help you engage with the coursework in your Ph.D program better, but will not be covered directly in that coursework. If you have one extra course to take before starting at Ph.D, my recommendation would probably be something closer to pure-math, like taking the calculus course you mentioned, a linear algebra course, maybe probability theory, etc.

Though this may not directly relate to the OP's situation, I'd say my advice shifts a little bit if you're thinking about this earlier and creating a plan of courses to optimally prepare you for starting a Ph.D. In this case, I think an ideal set of courses contains both more theoretical courses (e.g. pure math stuff) and more applied courses that can give you a better sense of how you'll use math in your research/can let you get started with research earlier. When I think of some of the best prepared (at least when it came to technical types) students to enter our program in the last couple of years, they tended to have done enough applied stuff to have a research agenda, but enough theoretical stuff to provide them with the "mathematical sophistication" Coach talked about. In my experience TAing one of the statistics courses in our sequence, I found that theoretical courses and general ability to math (yes, I'm using it as a verb now) were far better correlated with high performance in the class than was prior experience with applied statistics coursework (which, informally, seemed almost entirely uncorrelated with performance).

[catchermiscount]

Completely anecdotally, many of the people that have struggled with our technical classes are those with too many preconceived notions of what statistics and theory are. 

[RWBG]

Completely anecdotally, many of the people that have struggled with our technical classes are those with too many preconceived notions of what statistics and theory are. 

I suspect this is more of an issue with your program, as students who select in are likely to have thoughts about these things that persuaded them to choose your program over others. I think it's more common for those in other programs to come in with no notions on statistics and theory, preconceived or otherwise. 

[timeless90]

Hi RWBG and coachjrc. 

 

I think either of you are in a position of authority to answer my question!

so i guess my question basically trickles down to, "does the adcom have a preference for quantitative over the qualitative, and vice versa?"

 

Maybe Rochester is an exception since it is very well known for quant specific training, but in general, it is my impression that many of the successful candidates lack the quant background at the time they enroll in their Phd programs. In any case they have to relearn all the tools, stats and what not, from what i gather.  

 

But I found it ironic how many people at least on TGC are emphasizing the quant preparation in their SOP's.

I shouldnt generalize, but did these people simply happen to incorporate quant methods during their UGs, or are they leaning towards quant methodologies because adcoms tend to prefer those?

 

I did take upper econ, stats, multivariable calc and etc, but I am predominantly from an area studies background and I was wondering if I would be in a position of disadvantage for not having extensively discussed quant methodologies in my SOP

 

Thanks!! 

[catchermiscount]

I wish there was a hard-and-fast answer to the question.  There is not.  It very much depends on where you're applying and what sort of punchline that you're selling about yourself. 

 

You have far more technical background than I did the first time I applied to grad school, and part of your itchiness stems from a lack of ignorance yielding a lack of bliss.  Even if you didn't mention it explicitly or talk about how you might like to build upon it, people will notice that you have good background.

 

Many of the undergraduates here are exposed to quite a bit of quantitative and theoretical stuff---I am shocked to see what kind of things are done in honors theses.  But, my sense is that this is very much the exception to the rule.  And you're right:  while these students are laudable for doing interesting things with some amount of technique, they certainly don't know it at the level we would expect of a graduate student, so there is a lot of re-learning (or just learning in much more depth).

 

I've never been on an adcom before.  My sense is that, while it may be nice to demonstrate that you know what you're getting yourself into in terms of the technique in the writing sample/SoP, it's much more important to demonstrate that you have brain waves.  Can you be critical, can you find an interesting question, can you place it into a scholarly conversation?  Can you craft a compelling answer to the question, regardless of the tools you use to do so?  Can you consider rival explanations to your answer and argue in favor of your own?  I think that, if you did all of these things well, you could get away with the page numbers being the only quantitative stuff involved.