Advice on pre-PhD Quant/Stat Methods

[Zahar Berkut]

After a long hibernation, I'm emerging-- with questions for the informed!

I will have one free course next quarter in my area studies terminal MA program (terminating in June), and I'd like to take that chance to bolster my quantitative skills, which I currently lack. The issue is, I have many options available to me for a one-course venture into mathematical/statistical methods. My background: BC calculus from high school.

Previous posts on this issue suggest two philosophies: i) cover specific topics relevant to statistical social science work, or ii) just get the fundamentals of math down (calculus, differential equations, linear algebra, etc) that will allow someone to access the necessary coursework when the time comes. The trade-off is one of timing-- if you want to use quantitative methods on the job before pursuing a PhD, are there any topics that must be covered beforehand?

To help me decide, and especially to help me distinguish the different syllabi available to me, I ask the forum what topics need to be covered in order to:

1) Signal quantitative ability to ad-com's;

2) Signal quantitative competence to potential public/private sector employers;

3) Enhance my ability to conduct political science and/or policy-relevant research (two questions, I know);

The more specific one can be about topics that ought to be covered in methods, the better-- it will help me and whoever else faces a similar situation in the future compare syllabi. Currently I am looking at two stats-for-social scientists courses, one with a heavy R software emphasis and the other more theoretical; one statistics department course (still more theoretical, but it's hard for me to judge); and finally, a linear algebra/multivariate calculus course that itself would not contribute to research methods but would allow me to study econometrics and intermediate microeconomics on my own following the completion of my degree.

[brent09]

Theoretical stats for social science, learn R on the side. The background math is useful if you have tine, but given the time constraint just go for the stats class that will give you a good theoretical grounding.

[saltlakecity2012]

There's also the option of doing linear algebra. If you don't have a background in multivariable calculus, which at some point you should probably get, you can still do lin alg. whereas ODE would be pretty difficult, especially if you haven't done calc in a few years. It's a good signal, as it's not just number-crunching but rather also theoretical mathematics, proof-based. So it signals strength in logic as well.

Edit: Also, it's nice to see you again!

Edited March 2, 2012 by saltlakecity2012

[RWBG]

When you say theoretical stats, do you mean stats using a lot of calculus/preferably proofs? If so, it could be good, but only if you have a way to signal the content of the course (i.e. if the course title just says "Statistics I" it may not help, unless you have a letter writer talk about it). Linear algebra can be a bit more consistently theoretical (although yours is combined with multivariate calculus?) but the payoffs in terms of research applications are probably more long-term. I'd say linear algebra is likely to be best if you plan on doing quant once you start grad school, as most the stats stuff can be duplicated in your grad courses, while you may not get an easy opportunity to take a dedicated course in linear algebra once your start your Ph.D. Also, when you say methods, are you including formal theory or just stats?

[Jwnich1]

I would say get the fundamentals down. At the least cover 2 semesters worth of calc (basic differentiation and integration, taylor series etc. Nothing too fancy there), some linear algebra, some basic stats (linear regression!) etc. Which of these is better if you only have one course? probably any of the basics you have never touched. (if you have seen calculus before, take linear algebra if you havent). My view is that you can always relearn something faster when you need to, than learn something entirely new.

For the most part, you will take the more complex stuff once you get in to PhD programs. Most programs with the exception of perhaps the MOST quant - rigorous, know that poli sci training (at the BA level, and MA to some degree I'd imagine - don't know for sure) varies across schools from nearly STEM-like training to the old model of basically teaching the diplomatic history of the US. I wouldn't worry too much if you have a basic background in stats and math. Some formal methods might not hurt if you can find an undergrad course covering game theory in the econ dept (or poli sci if you're lucky) - grad courses work too, just dont know how in depth you want to go.

Edited: for severe lack of coffee

Edited March 2, 2012 by Jwnich1

[brent09]

Just to reiterate my comments (with a bit more discussion): I would advise taking the statistics for social scientists, preferably the course with a theoretical grounding. Linear algebra and calculus are important classes, but they don't directly apply to political methodology (they're nice to have but not absolutely necessary). You would be better off showing admissions committees that you can handle the statistics we use most frequently. If you had two years, then I might suggest going for all the background. But now you need to be strategic, and the best choice under those constraints is to focus on the stats you're most likely to use in a Ph.D. program.

Knowing R (and Stata, WinBUGS, JAGS, Amelia etc) is helpful, but it's perfectly possible to pick that up on your own. Take a good theoretical stats course and master the skills, and teach yourself some R on the side. In my opinion, that would make you a very strong applicant.

[saltlakecity2012]

Just to reiterate my comments (with a bit more discussion): I would advise taking the statistics for social scientists, preferably the course with a theoretical grounding. Linear algebra and calculus are important classes, but they don't directly apply to political methodology (they're nice to have but not absolutely necessary). You would be better off showing admissions committees that you can handle the statistics we use most frequently. If you had two years, then I might suggest going for all the background. But now you need to be strategic, and the best choice under those constraints is to focus on the stats you're most likely to use in a Ph.D. program.

Knowing R (and Stata, WinBUGS, JAGS, Amelia etc) is helpful, but it's perfectly possible to pick that up on your own. Take a good theoretical stats course and master the skills, and teach yourself some R on the side. In my opinion, that would make you a very strong applicant.

I have to say I agree that if you have no math background at the college level and you want to beef up your application to poli sci programs, stats classes would probably be more useful, especially if you have only 1 course. Linear Algebra is something you will take if you intend to do any formal theory (at least I hope), but if you're more interested in demonstrating that you're on board with the stats-heavy trend in poli sci, it won't be the strongest signal. But whatever class you do take, rock it.

Also, no one's mentioned this, but try to get your GRE quant score as high as possible. I don't know what you applied to your MA program with, but it can't hurt to have higher scores, especially in the quant section.

[iwouldpreferanonymity]

Since some people are suggesting an autodidactic approach, I wonder if we can establish a list of good online resources for independently learning math and stats. I know that there are some good materials available on iTunes U, and I have found Khan Academy to be helpful for some things. Perhaps there are some more obscure, yet equally beneficial, online offerings out there. Would anyone care to suggest some? (Yes, I could google for them - but I thought that raising the issue here might provide some benefit to others as well.)

[somanytictoc]

I've spent some time at AcademicEarth.org. There's a lot of overlap with iTunes U, but I don't have iTunes on my computer.

[Zahar Berkut]

Thank you for the responses!

There are some great insights here, and I'm thinking that if I have only one course on campus available, it should probably be the more rigorous stats course with an emphasis on applications (not the R-intensive one). I may still take the autodidact route and build up on linear algebra and other math course sequences, but the stats option may be most effective for signaling potential employers in the near term.

Personally, I'm not sure how I would be using quantitative training in a possible PhD program. I believe strongly in mixed methods, so any tools available to shed more light on a question is great, but I can't say right now how much I expect to work with formal models.

Come to think of it, that's probably a problem for the mixed methods philosophy: you can't have every possible tool available without committing a disproportional amount of time to methods at the expense of thematic courses. So the efficient methodology-training outcome probably won't move very far past some of the basic regressions.

[AuldReekie]

It's an interesting question. I know plenty of UK students make it into US PhD programs and they won't have done any math for at least three years!

[shavasana]

Who loves MIT Open Courseware?! ME! I literally re-learned calc in a couple of weeks with this! I use it inconjunction with my old calc book from high school.

[AuldReekie]

Who loves MIT Open Courseware?! ME! I literally re-learned calc in a couple of weeks with this! I use it inconjunction with my old calc book from high school.

Definitely agree. I'm currently teaching myself some basic game theory with UCLA's equivalent