Harvard Biostatistics v. Berkeley Statistics for PhD?

[2fluffydogs]

I'm deciding between Berkeley statistics and Harvard biostatistics for a PhD. I've been working as a biostatistician for three years, and want to focus on the more methodological areas of biostatistics. I'm not sure what I want to do after graduating, but want to stay in research.

Berkeley statistics offers a more theoretical program, but has relatively few faculty (~5) working on biostatistics. Moreover, those in biostatistics seem to be focusing on single cell data recently. But, the work is more methodological, and there is still some applied work being done. On the upside, Berkeley has no quals and the students seem nice. It seems like recent graduates are split between good positions in academia and tech companies. 

Harvard biostatistics is big, and there are a lot of faculty working on a ton of different things. The department isn't all applied, and seems to have a bit of methodological work (including some Bayesian stuff). There're also many hospitals in the area and the department has a ton of high quality data. It seems like recent graduates end up post-doc-ing a bit, or going to biostatistical research institutes, but that post-docs from Harvard tend to get good positions in academia. 

My SO might be able to transfer his job to the Bay area, which is an advantage, as is the better weather. Cost of living isn't that different between both places. Harvard is ranked #1 for biostatistics, and Berkeley #2 for statistics (though note that Berkeley biostatistics is #6, and the biostatistics and statistics faculty overlap). 

Is anyone familiar with the programs, or with the more intangible factors from both departments (e.g., culture)? I welcome any thoughts that anyone has!
 

[stat_guy]

I'm just another stats/biostats PhD applicant this year. I've been working at Harvard biostats for a while. IMO this is a more applied department, since as you know they have arguably the best medical school in the world therefore many high-quality data. But they do have some theoretical people, like Rajarshi Mukherjee, Tianxi Cai, Junwei Lu, etc.

Another factor you should consider is what topics in biostats you're interested in. For example, if you're interested in statistical genetics I'd vote for Harvard biostats, for example Xihong Lin is a pretty strong faculty in this area. If you're interested in causal inference, Harvard is good at it too but I personally favor Mark at Berkeley. Overall I agree there are more people working on theory and methodology at Berkeley stats than Harvard biostats. You can't go wrong either way, best of luck to your decision!

[statsnow]

I think Harvard is a much better biostats program than Berkeley biostats.  I think Berkeley stats is a better program than Harvard stats.  If you are dead set on biostats go to Harvard.  However getting a degree in stats gives you more flexibility.  Depending on what you want to do, going to Berkeley gives you more options for the future.

[2fluffydogs]

Thanks for your thoughts! It's definitely a hard choice to make. I think that I'm definitely interested in biostatistics or at least statistics for biological applications.

Harvard has more and better data, more faculty, and more options, but the program has more requirements and is less flexible overall. The first year coursework is much easier than Berkeley stats', but there's quals. Looking closely at placements of people who worked with my advisors of interest (while they were at Harvard at least), they seem to be various research scientist positions at big centers or industry. The students at Harvard tend to come from broader backgrounds, while those at Berkeley stats tend to have math undergrads.

Berkeley has fewer faculty and less data, but it also has some well known people working in the area--Dudoit, Bin Yu, Purdom, Huang, van der Laan. The first year coursework sounds difficult. A lot of placements of phd grads from Berkeley stats working in biostats are in industry, with a few postdocs here and there (and, there's been a few Berkeley stat/biostats phd grads who end up postdoc-ing at Harvard biostats). There's a lot of good methodological work, as well as more of the high dimensional stats flavor to the work at Berkeley; however, a lot of people there are primarily looking at single cell data.

Almost everyone I know says Harvard (which makes me wonder if I'm overthinking this or if it's obvious), though my SO's easy job transfer makes Berkeley a nice option.

[bayessays]

I don't think there is an obvious choice for sure, but it probably should come down to personal factors that nobody else besides you will be able to decide.  Personally, it sounds to me like academically you'd be happier at Harvard - it will satisfy your desire to do methodological work, you'll have a ton of options, etc.  Harvard's quals only test first-year material, so I'm guessing they wouldn't be too horrible, but you will know yourself better than the forum - will these tests stress you out so much for a year that it will hinder your quality of life?  It only takes one advisor to get a PhD, so I wouldn't be too worried about the options at Berkeley if you can find some faculty you like *and* are fairly confident they will take you as a student.  There is also somewhat of a cultural difference between stat and biostat programs -- I imagine Berkeley has a lot of people who are just in interested in math/machine learning, and at Harvard you will actually be at a school of public health and many people will be interested in drug research, vaccines, epidemiology, etc.  It's just a different social environment. As for jobs afterwards, you're going to have a PhD from a top stat/biostat program at the end anyways and you'll be able to get many many satisfying jobs, so I don't think you should worry too much about this.  I would personally weigh the personal factors most heavily (SO's job, location preferences, whether you will be personally stressed by quals) rather than the academic/prestige/career concerns.

[PLessPoint05]

As an aside on the cost of living, having lived in Cambridge for two years and now Bay Area for over 3 years, I would say Cambridge is a bit lower. 

[StatsG0d]

I think Berkeley is the clear winner if one of following apply:

1. You are really interested in causal inference
2. You care a lot / enjoy theory

Otherwise, I'd say go wherever you feel more comfortable / interested. Personally, I like Berkeley's model (no quals, customizable coursework, etc.) 

[bayessays]

1 hour ago, StatsG0d said:

You are really interested in causal inference

Isn't Harvard big in causal inference with van der Weele, Robbins? Just curious, I know van der Laan is big too but I thought causal inference was a big strength of Harvard (they also have Imai in the stats department).

[statsnow]

Rubin retired and went to china so Harvard has lost a lot of its causal inference focus

[StatsG0d]

3 minutes ago, bayessays said:

Isn't Harvard big in causal inference with van der Weele, Robbins? Just curious, I know van der Laan is big too but I thought causal inference was a big strength of Harvard (they also have Imai in the stats department).

For sure. I was unclear. What I meant is if you're sure you want to do causal inference, go to Berkeley because it's a central focus of their department (particularly biostats) and you get the advantage of the culture / higher rank etc. 

As a side note, Robins is great, but he still does very traditional causal stuff (IPW, instrumental variables, etc.), while van der Laan is focused on very modern stuff (lasso, nonparametrics). Also, the primary appointment of Robins is in epidemiology, so it's not clear to me if he's generally available. He's also in his 70s, and I can't imagine he sticks around *too* much longer. That said, he still regularly publishes in JASA, Biometrika, etc.

[Blain Waan]

@2fluffydogs Do you know who's the chair of this year's Biostatistics PhD admission committee at Harvard?

[2fluffydogs]

@Blain Waan Sorry, no idea who the chair is.

[causalinf]

Paige Williams is currently the Director of Graduate Studies. She will step down soon and Sebastien Haneuse will take over to become the new DGS.