Jump to content
Urch Forums

gollin

Members
  • Posts

    8
  • Joined

Converted

  • My Tests
    No

gollin's Achievements

Newbie

Newbie (1/14)

1

Reputation

  1. If you do measure theoretic probability you could simply read Casella and Berger afterwards and learn everything. Analysis is harder to learn on your own as many books eschew motivation in favour of a definition-theorem-proof type of exposition although analysis is fundamentally an intuitive subject (the rigour is just needed to refine your intuition initially). A class can really help you develop that intuitive feel for the subject, both through the broader insights of the professor and from your peers who you would be solving problems with. This is useful EVEN if you have a sufficiently well developed intuition for undergraduate analysis
  2. Another thing that I forgot to add was that you should try to find projects where your would have an absolute advantage —vis-a-vis a comparative advantage— in some aspect of the project. Now be careful because the average research economist is going to be better than you at almost all parts of the project they work on; they just choose to outsource some tasks because there are usually better uses of their time. But some of the newer projects require skills that many economists don’t possess, particularly in relation to management of large relational databases, running code on distributed clusters (many universities have Hadoop/Spark crash courses offered by the HPC people), applying machine learning techniques. On the last point, many economists are familiar with machine learning in the perspective of an econometrician and generally think of them as a class of estimators that do well as out of sample predictors. As such, they’d be familiar with machine learning in the context of inference (so things like double selection Lasso.) However, many would not know how to use specific tools well known in the ML community such as convolutional neural networks (or would not be familiar with the standard frameworks that implement them). I know someone whose choice of machine learning method helped improve the power of his PI’s instruments enough to get his name on the paper, and himself into a top 5 program. Obviously, economists are savvy people so this gap will quickly closed (and many top economists are already pretty well informed on these subjects) so better hurry and grab these opportunities as they come up!
  3. laborsabre, you are exactly correct! We are (going to be) economists and not mathematicians (and most of us tend to understand the principle of comparative advantage). The majority of the matriculants at a PhD program would become applied empiricists who would perhaps be better off reading good empirical papers rather than reading proofs. It is just strange that we sort of go half-way and make things rigorous so that the applied guys feel like they are learning a lot of esoterica and the theorists are left dissatisfied, rather than making theory an elective which would present the material by building up all the tools necessary all the while getting the applied guys learn some of the same things but in a more applied research context, mostly omitting proofs completely unless they serve some pedagogical purpose. (By the way Brouwer is not esoteric at all and is fundamental to topology and characterizing the behavior of continuous functions on sets which are homeomorphic to a compact Euclidean ball and Kakutani is the set valued generalization) I think I am maybe deviating into a whole new debate here, but I have frequently been perplexed by the way the Econ PhD is structured. My interests in math are pretty much orthogonal to Econ and the math I know hardly helps me better understand the work I do and the papers I read for my job. I don't see, therefore, the need for such a rigid core. People doing CS PhDs usually take courses as they need them for research; this is true at many math departments too. If something in my research would require some heavy theoretical machinery I would try to get a competent coauthor who understands it well rather than try to naively attempt to do everything myself.
  4. I think I would like to elaborate on a couple more points, in-line with the spirit of masstech's comments. If you trying to become an RA to boost your profile pick your program/advisor carefully: Often, people neglect to choose their RA advisors as carefully as they would pick PhD advisors thinking that as long as their advisors are at a top 10 school they should be fine. This is NOT fine. Firstly, try to pick projects which you have some baseline interest in. No brainer. This would allow you to be actually passionate about the work you do and would make a good impression. As a corollary, don't just fire off hundreds of applications to every posting on the NBER RA page. I know the marginal cost appears to be negligible but you do not want to end up at a position where your interests don't align with your advisors. I had such a position at a school and then moved to another school because of I thought I could make better contributions in a different field. Secondly, try to evaluate the placements of your advisor's previous RAs. This is often highly correlated with how willing these professors are to leverage their connections to help you. Some professors don't care about placements in any shape or form and just want your labour. Some professors care about the placements of their PhD students but don't really think about their RAs. Some professors work really hard to place their RAs (Chetty, Gentskow/Shapiro are obvious examples but there are less famous examples like Zack Cooper, Johannes Stroebel). Three, try to join one of those structured predoctoral fellowships if you can. These include Opportunity Insights (probably the most selective), Stanford (SIEPR), MIT (SEII), Yale (Tobin), Chicago (EPIC, Heckman) etc. These are either big, well organized labs, or are formal programs (with coursework requirements etc) in which the sponsoring departments have a vested interest in making good placements. You have often get the added benefit of dedicated program director who can help with admissions advice. Of course, within these fellowships you should try to pick topics in your interest area (see point above). Four, take graduate core courses if you can, although you should have a precise prior about your own abilities vis-a-vis the incoming class. Unlike excelling in, say, a real analysis class as an undergraduate, here the objective is to be better than other students. Obviously, this is really hard to know ex-ante so just ask your advisors once you get there. Lastly, try to avoid RA programs where you will not be close to your PI a. la JPAL/EPoD UNLESS that PI is well known for helping their RAs place well (I have heard Rohini Pande is a good "distance PI").
  5. Hi all, Thanks for your advice, I think I'm going to take Micro. Most of the other predocs are not taking this class but I think should be able to figure out(?) a study group. The professor teaching the second quarter of the class is someone I used to RA for at a different institution (we both moved here, coincidentally); hopefully I can get a letter out of this. My concern with this class (reflecting back on intermediate micro) was that while the material itself did not present to be particularly deep or challenging relative to the stuff in math courses, the exams were often deliberately tedious and time tight (lots of Lagrangians with unnecessarily complicated looking FOCs). I'd rather write proofs in a take home exam than sit through such an exam as I feel I am not a wizard calculator with calculus and algebra (unfortunately, my math professors drilled in me the notion that anything a computer could do better than me was not worth becoming particularly good at). Another unfortunate facet of the course which I have gleaned is that some of the material uses "big guns" which they don't explain; Brouwer's fixed point theorem, for example, is used for general equilibrium stuff but is never usually proved in Micro classes (the intermediate value theorem corollary for the one dimensional case doesn't count). I find this disconcerting because if the point is to be "rigorous" then "lemmafying" some of the biggest steps without explanation (and no one expects you to have taken a course in algebraic topology where this would have been taught) is a bit of a hack. I wonder what is the point of this, pedagogically? In no "rigorous" math course that I have done have we ever discussed results that require substantial machinery outside the scope of the course (except, perhaps, Levy's continuity theorem as a lemma in the proof of the central limit theorem; that was in a non measure-theoretic probability class and it was primarily a non-proof based class.)
  6. Hi dogbones, From what I have heard, doing a field course (in say development, which is what I work on as an RA) is not considered a positive signal since it is not very rigorous compared to the first-year core coursework.
  7. Hi all, I am RA at a top university in the U.S. and have the chance to enroll in graduate courses in the coming fall. I am split between taking the graduate real analysis course (measure and integration) in the math department or taking Micro I (consumer and producer theory, choice under uncertainty MWG chap 1-6). As a prospective economics Ph.D student, the obvious choice would appear to be Micro, given that it is a direct signal for performance in the first year coursework. However, given the pervasiveness of measure theory in various fields -- from statistics and econometrics to the economics of information and Bayesian learning -- it seems like a solid real analysis course may be good for my own edification. For context, I have been studying from Papa Rudin (RCA) for the past couple weeks, albeit progress is slow since I am doing all the problems by myself and have not really put all that much time into it given my other commitments as an RA. I suspect peer effects are important with respect to outcomes while engaging with material like this. I have no experience with MWG (only baby Varian). People here tell me that getting the highest grade is crucial in Micro but to do so you need to be in the top 5% of the class. That does not seem realistic for me given I will be competing with a host of international students who have seen the material before. I do not know how the grading works in the Math department. Does anyone have any advice for me? Did you take both the courses? If so, which one was more challenging and/or edifying? Appreciate the help
  8. Hey all, I am an RA at a top university and have the chance to take graduate coursework this coming fall. As a prospective economics graduate student, I am torn between taking the first semester Real Analysis course (measure and integration) with the Math Ph.D students vis-a-vis taking Micro 1 (Chapters 1-6 MWG). On one hand, the micro course is probably a direct signal of ability to perform well in graduate coursework, and for most people it would be an obvious choice. However, I feel measure theory- pervasive as it is in so many areas of thought including mathematics, statistics, econometrics and the economics of information- feels fundamentally more important to know and understand. Have any of you taken both courses? Which one felt more challenging and demanding? Which one did you personally feel like you learnt more from? I am currently working through Rudin's book Real and Complex Analysis. My progress is very slow, since I do problems by myself (hardly an advisable strategy if you want to make rapid progress in learning the material; I suspect peer effects are very prominent in courses like these) and I do not spend much time on it given my other commitments as an RA. I have virtually no experience with MWG and have only done baby Varian many years ago. People also tell me that I need to be in the top 5% of my class to get the highest grade in Micro which seems unfeasible as many graduate students (almost all international students) have seen the material before in masters programs, so then I would come in with a disadvantage relative to them. I do not know how the grading works in the Math department. Given all of this, do you have any advice?
×
×
  • Create New...