Physicist turning to statistics?

[daveyrocket]

Hi, I have a PhD in physics and I have decided I do not want to continue down the path of physics. So I am looking at alternatives and I think it would be very interesting to be doing statistics. Unfortunately, I never had any formal classes in stats, but most of my physics research has involved doing statistics-related things, like statistical mechanics and quantum mechanics (which is all about probability distributions). I've also done a fair amount of Monte Carlo simulations.

I was wondering if anyone has any tips on how to sell myself in statistics. Also, what are the best ways to search for jobs in stats? Are they more common in certain geographical areas?

 

[Link]

Wow. I find it surprising that you have a PhD in physics (a hard field) and not only want to get out of it, but into statistics (another hard field).

Are you planning on applying to the PhD program in Statistics? I worry that it'll be extremely difficult for you since you do not have formal classes in stats. I have a background and I still struggle ... especially when it comes to advanced theory and proofs (man I hate those).

A good way to get jobs in any field (IMO) is to join professional organizations. You could try universities too, since they have lots of companies trying to recruit there. You could also do what I'm doing (take on more than you think you can handle and hope for the best ... if you do handle it well then you'll be viewed as a commodity in the work force).

 

[daveyrocket]

No, I'm not planning on going back to school. I've had enough of school and I would like to get on with my life, which means having a job which pays a decent enough salary that I can buy a house, etc.

So what is it that you're doing? Can you talk about that a little more?

Edit: I should also mention, I'm currently a postdoc. I went to a career fair at my university, but there weren't any companies looking for anything to do with statistics.
Edit2: Also I am curious, you mention you're surprised. What would your expectation be? How are you surprised?

 

[Link]

well...I think you've come to the right field. With the explosion of number crunching computers in the last few decades, its become more and more easy to not only collect data but to model and analyze it. There are so many things that you can do with stats that I don't even know where to start.

One conference I attended recently was this. Read through it. It should give you a really good summary of just what stats can do.

One piece of advice that I'd like to throw out there though is that you won't be valuable (i.e. won't be paid a lot) unless you know what you're doing. You'll have to learn how to program statistically using either SAS or R (or both). They're the most common programs I've noticed. In addition, you'll have to learn how collect data....to manage, clean, analyze, and model CORRECTLY ... how to make inferences based on your conclusions, and how to present your findings to lay persons.

PS. I'm in the biostatistics and epidemiology field. A one sentence summary would be: I use statistics to help quantify the effects of diseases, treatments, and to determine differences between certain kinds of groups and populations.

 

[daveyrocket]

Yeah I don't have experience with R or SAS, but I do have experience with Matlab, Octave and programming in general (C/C++,VB,C#,Fortran,etc.) as well as relational databases.

Of course things like collecting data and analyzing it show up in physics. What type of modeling are you referring to? Things like regression analysis, microscopic models, or solving differential equations given appropriate initial conditions, etc. come to my mind off hand. Is that the sort of thing you're referring to?

 

[Link]

Hi there,

surprised? well, usually people get out of hard fields cause its too hard for them. Some excuse is usually made about not wanting to put in the work or whatever, but they usually end up going into easier fields. I see both physics and statistics as hard fields. The better you get at it, the harder it gets.

Anyways, you might be able to get by then with Matlab and Octave ... but I'd still learn SAS and R if you want to become more valuable. As for the modelling, I'm not referring to anything specifically. It's problem specific. Whatever question you're trying to answer with whatever data you have on hand, that should guide you in choosing and implementing your modelling.

Good luck in your choice though. You should stick around ... try and see who you can help here. It'll help you become more fluent in this subject.

 

[daveyrocket]

Yeah the difficulty of stats doesn't really bother me. As far as leaving physics, the things that concern me are that the job prospects are poor, I'm not interested in the work anymore, and I have some geographic constraints I'd like to stick to, and physics is rather poor for that. I could go on, but I don't want to turn this into a rant.

 

[mads_st]

If you want i bet you could get a job in finance modeling complex derivatives and stuff like that. A lot of engineers, physicists and programmers work in finance. They do all the stuff us economists don't have the math and programming background to do
Also, if you program C and Matlab I see absolutely no reason for using more high level statistics tools. I've tried a lot of different programs and now i use Matlab for everything except perhaps a few things which i do in Excel.

 

[ichbin]

I am an ex-physicist (Ph.D. and postdocs in elementary particle theory) and now work as a computer programmer who does a lot of statistics. I also didn't leave because it was "too hard": I loved the work, but not the low pay and the academic ladder-climbing. Here's my advice.

First, look for statistics-heavy computer-heavy jobs, not jobs that actually specify "statistician". Quants/actuaries and bio-informatics are examples of fields where there is a history of hiring math-savy, computer-savy people without requiring any particular degree. Biostatistics and academic statistics are examples of fields that do not.

Second, while you don't have to go back to grad school, you do need to be prepared and interested to learn a lot of statistical theory. Yeah, you probably know how to do error-weighted least squares and understand null hypotheses and Bayes' theorem better than all the social scientists you talk to. But you probably don't know what a Mann-Whitney test is or how to do an ANOVA, much less more theoretical stuff about the Cramer-Rao lower bound on estimator efficiency, etc. Grab some good, theory-heavy statistics books (not anything titled "Statistics for [insert field here]") and dig in.

Physicists are smart and can learn a new field quickly, but they also have a tendendy to think that they needn't bother learning a new field in order to contribute to it, and that's usually wrong.

 

[Dason]

Physicists are smart and can learn a new field quickly, but they also have a tendendy to think that they needn't bother learning a new field in order to contribute to it, and that's usually wrong.

I tend to agree. I think Link underestimates the amount of math physicist have. For a physicist I wouldn't worry about the theory. It would be the applied side of things that one would need some background on.

Also: this.

 

[daveyrocket]

ichbin, can you recommend any books? As a side note, I picked up a copy of the cartoon guide to statistics at a used bookstore a few months ago... a fun read but not much more.

I certainly don't have a problem learning new stuff to contribute to a new field, but when I consider various career options, I only have so much time to study. Ideally I could get a job and move forward from there, but I have yet to see how realistic that is (different employers I've talked to have had different attitudes about that).

 

[ichbin]

There is a discussion of good statistics books here. The Cramer book that I recommend there is an old classic, the Casella and Berger book that several others mention is a more recent graduate-level survey.

For specific applications, other books are relevant. The classic of quantitative finance is Hull's "Options, Futures, and other Derivatives". "Quantative Risk Management" by McNeil et al. isn't a classic, but it taught me risk analysis and copulas. Other people can chime in with good texts specific to their sub-field.

 

[cruzeconomics]

I think your best bet for finding a good (read: high paying) job with minimal cross-over cost is probably something in finance. The majority of tools that economists and people in finance use are the same as those used in physics (we like to steal equations from you guys). If you have a physics PhD from a relatively well known program, then you shouldn't have too much trouble finding a job on wall st.

 

[Pandoga]

Yeah. That's probably the best idea. Hedge funds for example are doing a lot of number crunching these days and they are hiring people from the hard sciences to do that. Very-well paid after some years of experience.

 

[daveyrocket]

Yeah unfortunately wall st. is very far outside my geographical restrictions. Besides, I would hate living in New York.