Program Rankings vs Fit (Biostat vs. Stat)

[bayessays]

One decision that I (and presumably others) will have to make soon is whether to attend one of the biostatistics or statistics department I've been admitted to.  How big should the stat vs biostat difference make in our decisions?  For instance, I've been admitted to top 5 biostatistics departments.  Would it be ridiculous to attend a USNews 30-40 ranked statistics program instead?  If one wants to work solely on theory/methodology (finds consulting work incredibly boring and doesn't necessarily have an applied research interest), would the statistics department be the better choice?

[Tronathan]

The USNews list is confusing. UCLA, UCDavis, and Columbia are all in the 30-40 range and many of the rankings are Biostat departments. From what I have seen the top 5, maybe top 10 biostat programs have a lot of faculty that publish on methodology. If you have access to journals look at the publications of the faculty that specialize in the topics you are interested in.  

[DMX]

Depends on which programs (e.g. is it Minnesota biostat vs Yale stat? or Harvard biostat vs UF stat).

 

If you want to work soley on theory (though why wouldn't you want to do both? ), I think stat departments make more sense. Biostatistics (almost by definition) is applications of statistics to biomedical problems.

[cetaphil]

if you are 100% certain you want to apply stats to biological data then you should go to a biostat department. stat departments can be applied as well, but they usually have a much broader range of applications (astro, geo, neuro, ecology, finance, ML, sociology....) as well as being a bit more theory focused. Although at the top biostat programs you will also get a hefty dose of theory. 

[Shostakovich]

I think at a top 5 Biostats department you have access to enough theory work where you would be eligible for academic positions at decent Stats departments as well. The downside (or the upside depending on how you look at it personally) is that you have to specialize in certain parts of statistical theory, so if you're indifferent to that I'd go with the Biostats department.

[JZappa]

Shostakovich, I'm not sure this is true. I know at UNC our Biostats department is severely theory deprived compared to our regular stats department. I know sever biostats students and from what I see of their work and what they say they are doing the theory is almost nonexistent compared to what one would experience in the statistics department as a PhD candidate. For example I'm not sure they even do a heavy duty treatment of measure theory or measure theoretic probability. Both of these things are par for the course in a regular statistics program.

 

Edit: I stand corrected the PhD track does require measure theoretic probability but it does seem to be much slower paced than what I've experienced in my measure theory class.

Edited February 12, 2013 by JZappa

[health_quant]

The level of theory seems to vary quite a bit across Biostats departments.

 

For Michigan, the Biostats doctoral students take the same requisite graduate, upper-division sequence in mathematical statistics as Stats doctoral students, but are not required to take the measure-theoretic course in probability. Perhaps it's also worth noting that Michigan's more measure-theoretic coursework is cross-listed in the Math department.

 

Similarly, Penn's measure-theoretic probability sequence (with stochastic processes) is cross-listed under the Math dept. The doctoral-level stats course in mathematical statistics is taken concurrently with measure-theoretic probability, so it seems unlikely that their mathematical stats work(at least at that level) requires measure theory. Penn's Biostats department doesn't require measure theory (as it has its own sequence in probability and mathematical statistics), but evidently, many of the theoretically-inclined Biostats students take the aforementioned probability sequence through the Stats/Math department as electives.

 

The University of Washington's probability and mathematical stats courses are the same for the Stats and Biostats departments. From what I understand, students from both programs take the same theoretical sequence for the first two years. (I believe UWashington is known for being one of the more rigorous biostats programs.)

 

Wisconsin-Madison's Biostats program is just a concentration within Stats. In this case, the probability and math stats coursework are the same, i.e., lots of measure theory. 

 

I would imagine that biostatisticians with strong theoretical backgrounds (from the stronger programs) might still be viable in stats departments, but given the number of Biostats departments across the country (and industry positions), I don't think there's much reason for them to pursue those jobs. When you flip through the top stats journals (e.g., JASA, JRSSB), you'll see an abundance of biostatisticians along with statisticians. For us, the distinction seems to arise more from our relative emphasis on biomedical research applications.

 

Even with Stats departments, research in cutting-edge methodology (which of course requires extremely strong theoretical backgrounds) is gaining in importance. We can see that with how the University of Chicago's been updating its Stats department, and with the number of hires from other backgrounds (especially EECS) for machine/statistical learning, and topics in non-parametric Bayes. Really, it seems like much of the disciplinary boundaries between Biostats, Stats, and CS are blurring now more than ever. More traditional topics in theory, (e.g., probability and stochastic processes) still seem as much the domain of mathematicians and applied mathematicians as of statisticians.

[Shostakovich]

Shostakovich, I'm not sure this is true. I know at UNC our Biostats department is severely theory deprived compared to our regular stats department. I know sever biostats students and from what I see of their work and what they say they are doing the theory is almost nonexistent compared to what one would experience in the statistics department as a PhD candidate. For example I'm not sure they even do a heavy duty treatment of measure theory or measure theoretic probability. Both of these things are par for the course in a regular statistics program.

 

Edit: I stand corrected the PhD track does require measure theoretic probability but it does seem to be much slower paced than what I've experienced in my measure theory class.

 

UNC may be an exception to what I said. I did see that they were heavily on the applied side when they sent me an invitation to their visit day later this month.

 

At my undergrad we had a pretty small Stats department of around 10 faculty members, and I know at least two of them had PhD's in Biostats.