PDA

View Full Version : Measure theory



fp3690
02-17-2008, 05:34 PM
I know it's kind of off topic now with admissions pilling up, but I need info urgently. I've been kind of auditing a class on measure and probability theory this term (which, by the way, is in another ballpark of rigor than analysis). Because I don't get it for credit, my incentive to study for it is small, and the opportunity cost quite high, as I can instead revise beforehand for calculus iv and analysis.

My point is: how much do I really need measure theory for admission? Is it not something I can pick up while in the PhD? One of the people taking the same class was my econometrics TA, who has already finished his PhD coursework requirement, and said he SORT OF needs measure theory for his thesis!

Thanks and best of luck to all of you refreshing admissions pages every 5 minutes!

Olm
02-17-2008, 05:38 PM
When you say admission, do you mean getting in or succeeding in the program?

If by getting in, no, you don't need it. Just having analysis is fine.
If by succeeding in the program, more math certainly helps, but there are a lot of schools that make you take mathematical economics courses during the first year that cover measure theory/topology/etc. topics so you'll learn them anyway.

pevdoki1
02-17-2008, 05:41 PM
I would tell you not to kill yourself with too many math classes.

I think we're covering the Lebesgue integration chapter in Rudin in my analysis II class this semester, and I'm not worried about not having a full on course on measure theory. I really don't think any PhD coursework expects you to be familiar with it.

fp3690
02-17-2008, 05:42 PM
Thanks, that's exactly the kind of answer I was hoping for. I mean getting into (I'm applying next year). As for succeeding in the program, I think that I can pick stuff up on my own, since I won't really need to prove anything, just be able to read the theory.

It would be nice to have someone doing a PhD now to answer my question though.

Thesus
02-17-2008, 05:44 PM
This is a prof who has an AER publication:

"Unless you're planning to be a really hard-core math jock, measure theory is a waste of time for economists. I sat in on a course during grad school and took copious notes - because that's what you do in grad school.

Those notes are still in my office, some twenty years later, and I've never had occasion to look at them. It turns out that if spend you life in R^n, all that stuff just doesn't matter."

Olm
02-17-2008, 05:44 PM
All of the courses in my Master's program were PhD courses. You will be fine.

polkaparty
02-17-2008, 05:50 PM
"It turns out that if spend you life in R^n, all that stuff just doesn't matter."

I heartily disagree with this but it makes it better for me as there is some sort of arbitrage opportunity!

The reason it's useful to generalize is that you can isolate the key properties driving results. R^n is loaded with various properties of all sorts. If you don't worry about generalizing, it will be more difficult to figure out which one is crucial to your proof. Many economists know a lot of math, but it appears that few think like a mathematician.

Ultimately, I don't think any economist needs to worry about this stuff unless they are going into heavy theory, and that's very few.

asianecon
02-17-2008, 05:56 PM
I heartily disagree with this but it makes it better for me as there is some sort of arbitrage opportunity!

The reason it's useful to generalize is that you can isolate the key properties driving results. R^n is loaded with various properties of all sorts. If you don't worry about generalizing, it will be more difficult to figure out which one is crucial to your proof. Many economists know a lot of math, but it appears that few think like a mathematician.

Ultimately, I don't think any economist needs to worry about this stuff unless they are going into heavy theory, and that's very few.

In the first place, Thesus' prof publishes in AER-type journals and what polka is saying might not be that relevant or even publishable in these journals (the layout of AER itself is not that friendly to mathematical writing!). They're more interested in the economics that drives the result and not the properties of the mathematical model.

Publishing in more technical journals is a different matter. :)


Btw, I had a discussion with an econometrician at a top 1/2 program and he asked me what I think I lack which I replied I don't have measure theory. He then said that while it can be sometimes helpful to appreciate things, taking a year-long math class in it isn't that productive.

Ancalagon The Black
02-17-2008, 05:59 PM
I have done Measure Theory in my Real Analysis class. I didn't find it that difficult. I would say, the more math the merrier but don't kill yourself for it. Just try and get the best grades with what you have and if you think that you cannot match up to something, then don't.

Don't screw up your transcript to take a class when it is really not an absolute requirement. Try to get all A's in the other one.

PS: What is auditing a class?

fp3690
02-17-2008, 06:01 PM
PS: What is auditing a class?
Observing lectures without taking the course for credit.

polkaparty
02-17-2008, 06:18 PM
In the first place, Thesus' prof publishes in AER-type journals and what polka is saying might not be that relevant or even publishable in these journals (the layout of AER itself is not that friendly to mathematical writing!). They're more interested in the economics that drives the result and not the properties of the mathematical model.

The two go hand in hand, that is, the properties of the model and the economics behind it. The reason they aren't currently 1-1 is exactly because of economists' lax attitude towards precision. I understand that most disagree with my views, but I just wanted to throw them out there.

Yes, the AER's two column per page is painful, for any kind of writing.

Olm
02-17-2008, 06:29 PM
I like two column formats :( so much easier to read, less wasted space (given the ~60 characters per line limit)!

reason
02-17-2008, 07:15 PM
Don't screw up your transcript to take a class when it is really not an absolute requirement. Try to get all A's in the other one.


This may be material for a another thread, but which is better: getting all A's in the degree subjects, or getting all A's in the degree subjects and B's to A's in additional* math courses?

*Additional as in taking more courses than required for the relevant degree.

macroeconomicus
02-18-2008, 02:34 AM
You don't need to use measure theory during the first year. In fact, you probably don't HAVE to know measure theory for any other class that you take. However, if you find out that you need to know measure theory or measure theoretic probability for your research, then you could just take the graduate course at the math department if and when you need it. Most departments allow you to squeeze in a free elective course from some other department after the first year.

Ancalagon The Black
02-18-2008, 04:00 AM
To everyone who knows:

Since i learnt from fp3690 what auditing a class actually means, I would like to ask everyone else what the point of auditing a class is? Why would you sit and do a class if you don't get credit for it? Because at the end of the day, credit matters in admissions no? Unless the adcoms consider auditing a class also?

reason:

I'd say, leave nothing to chance and get good marks in all the courses taken !! :D But others can probably have a better viewpoint than I have.

phdphd
02-18-2008, 04:36 AM
I would take measure theory not because of the subject per se but to signal to the adcoms how profficient I'm dealing with abstract material. This is the way to distinguish yourself among the pool, right?

I think auditing a class is better than nothing, but in the end of the day is like reading a math book without solving the exercizes, that's to say, it doesn't count that much.

By the way, our old friend Tom has some suggestions regarding math classes

http://homepages.nyu.edu/~ts43/

econorama
02-18-2008, 06:28 AM
Originally posted by Ancalagon The Black:
Since i learnt from fp3690 what auditing a class actually means, I would like to ask everyone else what the point of auditing a class is? Why would you sit and do a class if you don't get credit for it? Because at the end of the day, credit matters in admissions no? Unless the adcoms consider auditing a class also?I don't know why fp3690 is auditing a class, but I can tell you why I'm auditing one. I got my bachelor's a few years ago, and while I took a lot of math courses as an undergraduate, most of it is stuff I don't use anymore. I work as a math tutor, so some of the material is still fresh in my mind. But a lot of it is never covered in my students' classes. One of my old professors is teaching a mathematical economics course which I'm auditing. I'm not getting charged for it; I just turn up every week and do all of the homework. Some of it is almost too basic for me, but it's useful as a refresher course. Plus some of it is stuff I really don't remember at all at this point. It won't be enough, and I fully intend to study some other books before I turn up at math boot camp at whatever school I end up at. But seeing my professor every week forces me to stay up-to-date in the class, which will get me through May. And by that time, I will have either quit my job or cut down severely on my hours to give myself more time to study on my own.

I probably could've taken the course for credit, but couldn't justify spending that kind of money when I may soon be paying for things like moving expenses so I can go to a good school. Unless a graduate program specifically asks my professor (he was one of my recommenders) none of them will know about the course I'm in, but it's more for me than it is for them. Math courses aren't just to impress admissions committee, they're so you don't drown in your first-year courses.

Ancalagon The Black
02-18-2008, 06:54 AM
econorama:

Very well phrased indeed !! I see your point of view. Unfortunately or fortunately, I have been brought up to think differently, probably from my background - getting into a good school in India is way tougher and even more selective than getting into even the best and most selective schools in the US.

Rarely do we do anything that does not help us to get admission. I see your point of view about auditing a class just so to help your get back some concepts and learn some more so that you do well in first year PhD. We don't generally think like that. We jump into (or are thrown into) the deep end of the pool and we scratch and bite and somehow latch on. :D

There is nothing like auditing in our education system (which is very rigid) Thank you for the information though. If I get into a school in the US, I am sure that there will come a time when I will also have to audit a class !!

snappythecrab
02-18-2008, 08:13 AM
You don't need to use measure theory during the first year.

That all depends on what you're doing. Students generally aren't expected to know it in their first year, but that doesn't mean that it won't come up in various, albeit small, capacities., eg: stochastic dynamic programming. It certainly doesn't hurt to have seen a bit of it.

asianecon
02-18-2008, 12:19 PM
econorama:

getting into a good school in India is way tougher and even more selective than getting into even the best and most selective schools in the US.


Sorry, off-topic, but I'm curious why you think so.

Sam_Jackson
02-18-2008, 01:51 PM
I think the main reason why the top Indian schools might be more selective is the sheer size of the Indian student body. If i'm not mistaken, the number of applicants to IIM-Bangalore is a 5-6 figure number.

A lot of my Chinese friends also say the same about their home universities (Tsinghua, etc). The sheer number of excellent applicants/students is probably the main reason why the competition is higher in these two nations.

This is just my take on it though; I could very well be wrong about this.

fp3690
02-18-2008, 02:12 PM
This makes sense. However, you also have to take into account that Indian and Chinese Universities cater almost exclusively (I think) to domestic applicants, unlike that of the US, where half grad students are international. Also the US are a 300m country with very high college enrollment (40%?), whereas India is a 1.2b country with, i guess, less than 20% college enrollment. If you crunch that stuff, it adds up to approximately the same market size.

buckykatt
02-18-2008, 03:36 PM
Also the US are a 300m country with very high college enrollment (40%?)

I'm not sure about total enrollment or 2-year degrees, but attainment of 4-year degrees is around 28%. Over half have at least some college, though, so if you're looking at the total size of the market I suppose it's somewhat higher.

Ancalagon The Black
02-18-2008, 04:34 PM
Well, to answer the curiosity of asianecon, let us take the example of my university Jadavpur University currently ranked 11th in the country in engineering in the latest rankings. 1000000 people apply via a one day test known as the JEEE which is just a state level test. 25000 receive a cut off for the various engineering colleges in the city and since my university is the top choice, the top 2000 get through to all the programs offered. So thats a selectivity rate of 2%

Now lets move even higher and consider the best 7 schools of the country in engineering which is the IITs (Indian Institute of Technologies) Around 1000000 people will apply via a one day test known as IIT-JEEE and around what 5000 will get through which is a selectivity rate of around 0.5 %

The reasons are:

1. Large number of students applying for too few positions. This leads to cutthroat competition where everyone studies as much as possible and learns up entire textbooks.

2. Entire selection process depends on a 1 day academic test.

Now comparing this with HBS PhD where 700 applicants apply for 21 positions. We think on a whole that its easier and much safer. Though those of us from non-IIT schools are jealous of the IITians. The Ivy Leagues et al love IITians. Safe choice. :( (eg. desimba is an IITian and I believe that he has made it to Wharton)

The same goes for the CAT test for admission to IIMs and GATE test for graduate students to IIT/ISI/IISC.

However, the number of applicants to an Indian university like IIT far exceeds the number of applicants to any US school (I could be mistaken but I don't think that I am)

asianecon
02-18-2008, 04:56 PM
I'm not convinced that looking at acceptance rates is the best measure of competitiveness (well, you don't really have to convince me :)) primarily because my guess is that the 700 guys/gals applying for HBS PhD belong to a highly selective group, compared to say the 1000000 people applying for Indian unis. Moreover, we don't really know the reason why there are only 700 people applying, ie if the cost of applying to Harvard and an Indian uni are the same(eg no application fees, no need for LORs, etc), 700 could well increase to say more than 1000000, so acceptance rates might not be comparable across different markets (ie Harvard arguably has a bigger market but at the same time higher barriers to participation/application).

Ancalagon The Black
02-18-2008, 05:05 PM
Good point asianecon but I would like to throw one more point into the fray. I don't really believe that most of the applicants to say, HBS are of the top category of students. Many, and believe me, many students with seemingly mediocre profiles apply to HBS just as a long shot.

Also, whilst I think that cost may be a factor for people applying in the US, in international applicants, it may not necessarily be true.

At the end of the day, however, I think that its just because we have a LOT of students. :D