Biostatistics/Public Health PhD instead of Statistics?

[fujigala]

TLDR: Does it make sense to apply for PhDs in Biostats/public health even if I want to do research in something not really related (machine learning/neural network theory), on the basis that admissions are less competitive and it would be advantageous to be at a better school, and one can still do research in such an unrelated area?

I'm currently a master's student in statistics looking to apply for PhD's in statistics next year to do research in machine learning theory (e.g., equivariant neural networks). My advisor in statistics recently came up with the suggestion to apply for PhD's in biostatistics/public health while still pursuing the same interests in machine learning theory (not really related to biostatistics/public health), since the admissions are apparently much easier. I would much appreciate any thoughts on the following:

• In a different field such as biostats/public health, is it still possible to do highly theoretical research in machine learning theory that is not really related to biostatistics or public health? My advisor seemed to think this was the case.
• If so, are there different ways to prepare for such programs? For example, for PhDs in statistics, taking a lot of heavy math courses is recommended.
• Is applying to such programs that are not exactly your primary interest advised or advised against? Biostatistics is one of my secondary interests, especially if it is hard to find careers in machine learning theory.

Thanks in advance

Edited November 6, 2022 by fujigala

[bayessays]

On 11/6/2022 at 11:58 AM, fujigala said:

• In a different field such as biostats/public health, is it still possible to do highly theoretical research in machine learning theory that is not really related to biostatistics or public health? My advisor seemed to think this was the case.

No, not really.   If you do a biostatistics PhD at a top 5 school, it may be possible to do some machine learning stuff, but it's certainly not going to be very theoretical.  What you should do is look at biostats PhD programs and go through every single professor's web page and see if they are doing anything that interests you.  Look at their papers in machine learning journals.  You will have to do the same thing for any stats programs you look at anyways, because ML research is still hard to find in most stats departments.

[fujigala]

@bayessays What I'm referring to specifically is maybe working with another professor outside of the department of biostatistics. For example, I know this is done in my statistics department where students will work with professors in other disciplines in addition to their own. But I don't know how acceptable it would be looked upon to do your dissertation in something completely unrelated to biostatistics. However, I do know that many master's students in my department will go onto PhDs in the business school, where professors still do highly theoretical research that is not directly related to business.

[bayessays]

A lot of business schools have professors doing statistics, yeah.  I know some pretty solid ML/stats people in business departments.  Psych, neuroscience, etc... you'll find people doing ML/stats.  Regardless, though, you have to find those people.  You absolutely cannot go into a program and expect that you'll be able to somehow get an advisor from another department.  I wouldn't go to a department that doesn't have at least a few people in the department itself that would be interesting to work with.

Part of this also depends on what you mean by "theoretical research."  What you consider theoretical might be more commonly thought of as methodology or even applied for some people.