Jump to content

Importance of program ranking for industry


Recommended Posts

Hi all, I'm in a bit of a bind trying to figure out my plans for next year. 

I applied for statistics PhDs this cycle and results were a bit disappointing: rejections from 5/6 of my schools and one funded acceptance from an institution ranked in the 30 - 40 range. I think my profile was reasonably solid (3.8+ GPA from a top undergrad, A's in real analysis and other advanced math courses, near perfect GRE scores, a few years of research) and I was a little surprised by the results -- per threads on here from past years, I got the vibe that I would be a competitive applicant at most schools.  

I like the school I got into a lot, but I'm worried that attending a lower ranked program will negatively impact my potential job opportunities down the line. I'm not interested in going into academia, but will attending a 30-40 ranked school put me at a major disadvantage compared to candidates from slightly higher-ranked institutions when it comes to getting industry research positions? And with that in mind, is there any merit to delaying grad school for a year to reapply? I wouldn't really be considering it, but I've heard that this year was more competitive than usual due to covid + I only applied to six schools, which seems to be on the lower end. 

Would appreciate any input you guys might have. 

Edited by 3musketeers
clarification
Link to comment
Share on other sites

No, it shouldn't put you at any disadvantage for industry. In fact, PhD grads from programs ranked 60-80 also can (and do) get jobs at Facebook, Google, and other "prestigious" companies. Industry as a whole is less "prestige"-driven than academia. Relevant work experience (e.g. an internship) and hacking skills are also more essential to getting most of these jobs than publications. I personally know Statistics PhD graduates who got jobs at Google and the like who had no publications.

FWIW, schools ranked 30-40 (and lower) are also good enough to land jobs at R1 universities and prestigious liberal arts colleges. You might not be able to get a job at Stanford, UC Berkeley, or an Ivy, but if you have a strong research record, you can still land a job at a solid research university. For example, last year, UNC-Chapel Hill STOR (a very solid program) hired somebody who has a PhD in Statistics from Florida State University. And North Carolina State University Statistics also hired somebody whose PhD is from University of Illinois-Urbana Champaign. 

Edited by Stat Assistant Professor
Link to comment
Share on other sites

I also accepted an offer from stat phd program ranked in the 30-40 range. To add some data for your reference, my alumni went to: US Census B., Battelle, Merck Pharm., Sandia National Lab., Apple, Westat, Lawrence Lab. UC Berkeley, Eli Lilly, Upstart Network, (postdoc: Columbia, Duke), Twitter, Amgen, IBM China, Google, and many other huge insurance and financial service providers. Currently, I'm interested in a stat faculty career and aware of fierce competitions these days in academia job markets; but industry placements seem to reflect much less weights on prestige.

Edited by statenth
Link to comment
Share on other sites

Agreed that it doesn't matter.  You'll get good jobs regardless.  I will say that some of the companies that are pickier about applicants though, and even a top PhD will not be a guarantee of a job.  I have seen people from top 20 programs interview at FAANG who didn't know the basics of logistic regression, and at some of the top companies you really have to ace the interviews.  But if you have a PhD from any stats program, and you are knowledgeable, I don't think any position is out of the question.  I'm not sure how many people really get "research" positions though.  I think Facebook has a job called "research scientist" that a lot of statisticians seem to get, but I'm not sure how different it is from being a "data scientist" at say, Google, where tons of statistics PhDs are just regular data scientists.

Link to comment
Share on other sites

I've worked in industry for my entire career, at a startup that eventually went public, at a well-known tech company, and now in R&D middle management at a more "old school" company near where I grew up.

Higher-ranked program helps a bit because (1) they're often doing research in hot areas that translates well to industry work (places like Stanford) (2) better connections when looking for an internship or job and (3) the "wow" factor of a big-name program - you'll get some extra attention, but you're not a shoe-in by any means. None of these are insurmountable if you come from a lower ranked school.

Biggest things that'll help in the job search are internships, dissertation topic, connections, and personality/likeability. If we're looking for a MS/PhD statistician to help an army of engineers with experimental design, statistical process control, and data analysis, we'd much rather take someone who did a lot of applied work at a school ranked #42 and had an internship at a place like Procter&Gamble, than take someone who spent their entire PhD working a theoretical topic in stochastic processes at school ranked #9. We've even hired Master's over PhDs if they were a better fit.

 

 

Link to comment
Share on other sites

23 hours ago, bayessays said:

Agreed that it doesn't matter.  You'll get good jobs regardless.  I will say that some of the companies that are pickier about applicants though, and even a top PhD will not be a guarantee of a job.  I have seen people from top 20 programs interview at FAANG who didn't know the basics of logistic regression, and at some of the top companies you really have to ace the interviews.  But if you have a PhD from any stats program, and you are knowledgeable, I don't think any position is out of the question.  I'm not sure how many people really get "research" positions though.  I think Facebook has a job called "research scientist" that a lot of statisticians seem to get, but I'm not sure how different it is from being a "data scientist" at say, Google, where tons of statistics PhDs are just regular data scientists.

@bayessays I have always wondered what kind of industry job really needs statisticians with a PhD degree. My impression is that a well-trained MS statistician (especially those with also an undergrad in Statistics) should be able to perform most jobs that the industry offers. What kind of industry job requires developing new theories, could you give me some examples? Also, is it common for companies to offer jobs where both MS and PhD statisticians can apply (I mean do they ever have to compete with each other)? Is it true that the starting salaries in the academic jobs are lower compared to the industry jobs (what about if the Assistant Professor is the co-PI of some grants, is the take-home salary still lower)? And how things look in comparison when the career progresses? Sorry to ask too many questions! Let me know if you think these questions should go to a new thread!  

Link to comment
Share on other sites

11 minutes ago, Blain Waan said:

@bayessays I have always wondered what kind of industry job really needs statisticians with a PhD degree. My impression is that a well-trained MS statistician (especially those with also an undergrad in Statistics) should be able to perform most jobs that the industry offers. What kind of industry job requires developing new theories, could you give me some examples? Also, is it common for companies to offer jobs where both MS and PhD statisticians can apply (I mean do they ever have to compete with each other)? Is it true that the starting salaries in the academic jobs are lower compared to the industry jobs (what about if the Assistant Professor is the co-PI of some grants, is the take-home salary still lower)? And how things look in comparison when the career progresses? Sorry to ask too many questions! Let me know if you think these questions should go to a new thread!  

Pharma has a ton of PhD-only positions. 

I went to a top-15 university, and while there were some very solid MS students, you simply cannot compare a <= 2year education that was mostly applied coursework to a >= 5year PhD that had rigorous theory classes, lots of applied classes and projects, and you had to know your stuff well enough to pass a difficult written qual, prelim oral exam, and had to solve an open problem and write 100+ pages about it. PhD students also had more time to master R, read literature, learn coding, attend conferences, TA/teach various courses, and had a greater depth of knowledge of the field in general. In some cases the PhD is also proof that someone can "stick it out", put up with a lot of BS, etc. as well. 

This is not a knock on MS students, and for some positions, we actually prefer the MS degree. We often have positions that accept both MS/PhD applicants and we've occasionally chosen an MS that was a better fit.

Edited by statsguy
Link to comment
Share on other sites

@Blain Waan

(Besides the pharma positions mentioned above, which definitely need a PhD, I'm mostly talking about tech here...)


I definitely agree that most jobs do not need a PhD.  However, I would say there are quite a few jobs where a PhD is significantly more likely to have the skills necessary, even if a well-trained MS statistician could do the job too.  At the end of the day, the student's PhD dissertation is likely not going to be the subject of their job anyways, so I don't really think a PhD is a true qualification for basically any job outside of bureaucratic requirements like the above-mentioned pharma case.  To answer your question about where theory is needed -- most statistics PhDs don't do super theoretical stuff and do more methodology, which is the skill that will often come in handy.

For instance, if you're a data scientist at a FAANG company, and you have to figure out what will be the best method to solve some problem, a random PhD student is much more likely to have the skills needed.  Most MS students essentially have a set of data analysis tools that they know and can apply.  But at a large scale, many practical and computational problems arise.  If we have data from billions of people that is too large to analyze in a standard way, and we only have access to some summary statistics on subsamples, how can perform statistically valid inference?   If only 60% of the data has come in because a subset of users has slow internet connections, how can we create a statistical method that takes this into account?  Most MS statisticians would have a hard time doing this by themselves and these are the types of problems data scientists work on at the big companies.

Link to comment
Share on other sites

On 4/16/2021 at 2:11 PM, statsguy said:

Pharma has a ton of PhD-only positions. 

I went to a top-15 university, and while there were some very solid MS students, you simply cannot compare a <= 2year education that was mostly applied coursework to a >= 5year PhD that had rigorous theory classes, lots of applied classes and projects, and you had to know your stuff well enough to pass a difficult written qual, prelim oral exam, and had to solve an open problem and write 100+ pages about it. PhD students also had more time to master R, read literature, learn coding, attend conferences, TA/teach various courses, and had a greater depth of knowledge of the field in general. In some cases the PhD is also proof that someone can "stick it out", put up with a lot of BS, etc. as well. 

This is not a knock on MS students, and for some positions, we actually prefer the MS degree. We often have positions that accept both MS/PhD applicants and we've occasionally chosen an MS that was a better fit.

Is there really a “need” though for the PhD in Pharma from an objective sense? A ton of the biostat work in pharma is not even actually statistical stuff, its mostly regulatory and documentation related. Pharma especially is super conservative and even uses SAS, which is the opposite of cutting edge. 

Having worked as an MS stat in biotech I hate it and that is why I’m looking at a PhD. But the PhDs Biostat I see in this area aren’t exactly doing the cutting edge work either. They are writing the same boring protocols and validation reports I am. Its not data analysis focused at all, as opposed to DS in tech. 

Link to comment
Share on other sites

20 hours ago, untzkatz said:

Is there really a “need” though for the PhD in Pharma from an objective sense? A ton of the biostat work in pharma is not even actually statistical stuff, its mostly regulatory and documentation related. Pharma especially is super conservative and even uses SAS, which is the opposite of cutting edge. 

Having worked as an MS stat in biotech I hate it and that is why I’m looking at a PhD. But the PhDs Biostat I see in this area aren’t exactly doing the cutting edge work either. They are writing the same boring protocols and validation reports I am. Its not data analysis focused at all, as opposed to DS in tech. 

It unfortunately doesn't matter whether the PhD is actually "needed" from a knowledge/proficiency point of view - pharma companies want PhDs whether we agree with it or not. Pharma has a lot of money to spend and very large R&D budgets, they are more than willing to pay a premium for PhD statisticians. Not only PhD statisticians, but they hire tons of PhD scientists as well for the senior R&D roles.

Also there are tons of roles in Pharma - I know one guy who works in optimization and logistics on the manufacturing side of a pharma company. Him and his crew of  statisticians and industrial engineers do lots of simulation work, use R/Python, and come across lots of "non standard" problems that a PhD would be well-equipped to solve. I know another guy with a PhD in Statistics who is a director of business analytics and data science at a smaller pharma company. Not everyone works on clinical trials, regulatory submissions, etc. 

Having been in industry my entire career (startup that later went public, then data science at a large tech company, now middle management at a manufacturing conglomerate) with a PhD from a top-15 department, the PhD has made life so, so much easier. I have seen a few people with Masters excel and move up, but there is definitely more of a ceiling. And on at least one occasion I've come across someone telling me that they want a PhD to fill a role because it simply "looks better". 

If you're young and don't have kids, it may be worth biting the bullet and doing the PhD. Good luck!

 

 

Link to comment
Share on other sites

20 hours ago, statsguy said:

It unfortunately doesn't matter whether the PhD is actually "needed" from a knowledge/proficiency point of view - pharma companies want PhDs whether we agree with it or not. Pharma has a lot of money to spend and very large R&D budgets, they are more than willing to pay a premium for PhD statisticians. Not only PhD statisticians, but they hire tons of PhD scientists as well for the senior R&D roles.

Also there are tons of roles in Pharma - I know one guy who works in optimization and logistics on the manufacturing side of a pharma company. Him and his crew of  statisticians and industrial engineers do lots of simulation work, use R/Python, and come across lots of "non standard" problems that a PhD would be well-equipped to solve. I know another guy with a PhD in Statistics who is a director of business analytics and data science at a smaller pharma company. Not everyone works on clinical trials, regulatory submissions, etc. 


What are some of the methods used in Pharma biostat outside clinical trials? 
 

And yea I am thinking of doing a PhD but I don’t know if I want to do Biostats again vs something like a DS PhD. It feels like the latter has a lot more interesting opportunities in industry, and seems to do more cutting edge biomedical data analysis, in big companies like Verily and Genentech. The branding is tough with Biostats I feel, because most jobs I see on LI are super formal SAPs, SAS, and the submissions.  I don’t even want to touch SAS again lol. 
 

Outside that it seems like there are omics jobs, but these need the domain knowledge and my program never covered them. So if I do Biostats I feel a good amount of Bioinformatics needs to be done too. I have gone for those positions but I lack the domain knowledge of all these sequencing technologies. 

Link to comment
Share on other sites

On 4/22/2021 at 10:26 PM, statsguy said:

And on at least one occasion I've come across someone telling me that they want a PhD to fill a role because it simply "looks better". 

Yeah, I got to the final stage of an internal transfer within a company to my dream location, and then the boss's boss wanted someone with a PhD.  I would definitely say just having a PhD versus a masters is a significantly bigger deal, in a variety of industries, than having a PhD from a high vs. low-ranked place.
 

16 hours ago, untzkatz said:

And yea I am thinking of doing a PhD but I don’t know if I want to do Biostats again vs something like a DS PhD.

Plenty of biostats PhDs go into data science, so you don't have to choose right away.  Really just think of a biostatistics PhD as the same as a statistics PhD with less options on the very theoretical/mathy end.  In fact, biostats PhDs often have a lot of good applied experience which is key for data science jobs, so I think a biostat PhD is a great option for that type of work!  But yes for something like genomics, you'll probably have to really specialize in the genetics end of things for your PhD.  But there are tons of data science jobs where a PhD in biostat would be helpful.

Link to comment
Share on other sites

1 hour ago, bayessays said:

Yeah, I got to the final stage of an internal transfer within a company to my dream location, and then the boss's boss wanted someone with a PhD.  I would definitely say just having a PhD versus a masters is a significantly bigger deal, in a variety of industries, than having a PhD from a high vs. low-ranked place.
 

Plenty of biostats PhDs go into data science, so you don't have to choose right away.  Really just think of a biostatistics PhD as the same as a statistics PhD with less options on the very theoretical/mathy end.  In fact, biostats PhDs often have a lot of good applied experience which is key for data science jobs, so I think a biostat PhD is a great option for that type of work!  But yes for something like genomics, you'll probably have to really specialize in the genetics end of things for your PhD.  But there are tons of data science jobs where a PhD in biostat would be helpful.

What about all the software skills though? Increasingly, data science is a field that requires the engineering aspects. You are competing with people who can make apps and put models into actual devices. Or dealing with big data technologies like Spark, Hadoop and cloud systems like AWS. 
 

This stuff is missing from Biostat curriculums. This is the aspect I feel my MS Biostat didn’t prep me for. I know statistical ML, but there aren’t many jobs in this area. May be though because I only have an MS, and one reason I want to do a PhD is to get those research scientist jobs that actually are stat ML focused. But even then I worry that there aren’t many of them and many PhDs end up needing to have the SWE skills anyways.
 

That and right now I am stuck in a boring biostat job which is largely documentation and writing/communication focused which I hate. The actual stats is simple. It seems my manager who is from biostat himself I guess is not even that interested in the stat aspects anymore these days. I want something more computationally heavy where I can use both advanced classical and ML methods and less writing/regulatory. 

Link to comment
Share on other sites

@untzkatz That is true, you don't learn a lot of the software skills, but if you know R you can pretty easily teach yourself Python, which would be a good career move, and you can learn SQL in a weekend.  As for the other skills, is there really a degree that is going to give people a lot of experience using things like Spark/Hadoop?  I learned how to run this stuff on-the-job, though I worked at a big company so they had more patience for learning things - a start-up would likely expect you to know a lot of that stuff.  You can learn a lot of this stuff on the internet, and you can choose a PhD department that will allow you to take electives in other departments to round out your skillset.  You can also do a dissertation on something that will allow you to pick up some of these skills.  For instance, I was a FAANG "data scientist" that was essentially a statistician, but now that I'm working on my PhD, I'm doing more of the machine-learning side of things, so I know if I wanted another data scientist job my options would be much broader.

I will say that you may be disappointed by the higher-level data science jobs too.  I am sure there are some places where you can do really exciting stuff, but a lot of business problems boil down to data cleaning and setting up pipelines, figuring out A/B tests, explaining them to people, etc.   But your mileage may vary, and these will still probably be much more exciting than a typical biostats MS job.

Link to comment
Share on other sites

4 hours ago, bayessays said:

 

I will say that you may be disappointed by the higher-level data science jobs too.  I am sure there are some places where you can do really exciting stuff, but a lot of business problems boil down to data cleaning and setting up pipelines, figuring out A/B tests, explaining them to people, etc.   But your mileage may vary, and these will still probably be much more exciting than a typical biostats MS job.

I know a decent amount of Python and am learning PyTorch myself right now doing a neuroimaging related DL project with a research lab as part of a class I am taking as a non degree student. I’m hoping this helps also for getting into PhD Bio/med informatics or DS programs. I still will apply to Biostat programs too but I don’t have the Real Analysis recommended requirement. Could take it this summer but I don’t know if I want to lol. I was always more interested in the applied computational side of things. I realized I really like the signal processing stuff and DL, and want to probably combine that with causal inference. 

Does a Biostat PhD still let you get into this computational side? I know places like UW with Dr. Witten (one of ISLR authors) does but UW is #1 lol and will virtually require Real Analysis. 
 

And yea I have been wondering how much actual stats is used in the corporate world. Some things I do today are so boring I am like why do I even need the MS for this, like a proportion confidence interval can be done by a high schooler who has taken AP stats and maybe knows the R function. I start to wonder if all this interesting ML/DL/AI and causal inference is even used much outside academia and FAANG. It seems like more than biotech companies, tech companies use more cutting edge methods. After dealing with the regulatory writing non-stat bullshit in biotech, I am less opposed to going to tech now. I like the idea of causal inference, AI in healthcare but its just not going to happen for a long time.

I’m sick of tabular experimental data too, there is less interesting statistically in a designed experiment vs the freedom of observational and unstructured data. 

Link to comment
Share on other sites

1 hour ago, untzkatz said:

Does a Biostat PhD still let you get into this computational side? I know places like UW with Dr. Witten (one of ISLR authors) does but UW is #1 lol and will virtually require Real Analysis. 

I would say Witten's research is pretty typical of stuff done at the top 10 biostat departments.

1 hour ago, untzkatz said:

And yea I have been wondering how much actual stats is used in the corporate world.

I also worked at a health insurance company briefly -- we did a whole bunch of causal inference/propensity score matching type of stuff, definitely not regulatory writing, so I think a lot of data science jobs, even in "boring" industries will give you some of the stimulation you're looking for.  Honestly, it was pretty similar to working at FAANG.

Link to comment
Share on other sites

19 minutes ago, bayessays said:

I would say Witten's research is pretty typical of stuff done at the top 10 biostat departments.

I also worked at a health insurance company briefly -- we did a whole bunch of causal inference/propensity score matching type of stuff, definitely not regulatory writing, so I think a lot of data science jobs, even in "boring" industries will give you some of the stimulation you're looking for.  Honestly, it was pretty similar to working at FAANG.

Yea DS seems to have that, but its kind of ironic to me that in the industry DS has more stats and tends to be more technical than bio”stats”. Fields that don’t have stat in their name like CS, EE, and within biotech-bioinformatics as well, all tend to use more advanced stats techniques in the industry than most biostats jobs. I kid you not I saw a Biostatistician job at an AI biotech company and then that position itself was all the regulatory, SAP, SAS stuff while the DS got the real stats. Its ridiculous.  Maybe a sample size or power calc or at most mixed model in Biostat. 

It’s quite ridiculous to me how many Biostat jobs want the PhD but then the description ends up having lame SAS stuff (to me SAS is hardly a stats software in 2021), FDA/ICH regulatory writing, and maybe just simple stats. I have seen people with PhDs who are in senior Biostat positions but then are just dealing with documentation all day. That is not real statistics.

I don’t know what it is honestly about the industry. I wanted to be a biostatistician, but many of those positions even at PhD level straight up don’t have as much stats at all. So I have to rebrand myself as a “data scientist” but this also involves learning more of the CS side, not just statistical ML. But that is for sure still better than writing. 
 

It certainly seems like Industry Biostat != Academic Biostat. Its an entirely different thing. They should label it something else because that sort of work imo (to me) dilutes the Biostatistician title. A field without the word -stat in the name shouldn’t be doing more actual stats than a field with the name. 

Edited by untzkatz
Link to comment
Share on other sites

I think it can matter for Finance companies if you're interested in quantitative finance (who typically only recruit from very prestigious universities), but otherwise I don't think it matters at all.

Link to comment
Share on other sites

On 4/24/2021 at 7:46 PM, untzkatz said:

Yea DS seems to have that, but its kind of ironic to me that in the industry DS has more stats and tends to be more technical than bio”stats”. Fields that don’t have stat in their name like CS, EE, and within biotech-bioinformatics as well, all tend to use more advanced stats techniques in the industry than most biostats jobs. I kid you not I saw a Biostatistician job at an AI biotech company and then that position itself was all the regulatory, SAP, SAS stuff while the DS got the real stats. Its ridiculous.  Maybe a sample size or power calc or at most mixed model in Biostat. 

It’s quite ridiculous to me how many Biostat jobs want the PhD but then the description ends up having lame SAS stuff (to me SAS is hardly a stats software in 2021), FDA/ICH regulatory writing, and maybe just simple stats. I have seen people with PhDs who are in senior Biostat positions but then are just dealing with documentation all day. That is not real statistics.

I don’t know what it is honestly about the industry. I wanted to be a biostatistician, but many of those positions even at PhD level straight up don’t have as much stats at all. So I have to rebrand myself as a “data scientist” but this also involves learning more of the CS side, not just statistical ML. But that is for sure still better than writing. 
 

It certainly seems like Industry Biostat != Academic Biostat. Its an entirely different thing. They should label it something else because that sort of work imo (to me) dilutes the Biostatistician title. A field without the word -stat in the name shouldn’t be doing more actual stats than a field with the name. 

You're overthinking this. How many positions have you had since you got your MS?

Just because you got stuck doing gruntwork in a biostat position doesn't mean industry is all SAS and one-sample t-tests.

When I was at a startup, our team worked around the clock to write algorithms to predict outcomes involving cancer diagnoses. Lots of R, Python, C/C++, etc. Based on what I later saw, most of the methods our team developed internally would blow away anything that was coming out of academia. But we were paid to help develop cancer screening methods that would ultimately save lives, not publish papers in journals whose audiences are 99% academics. Plus the whole thing about NDAs and confidentiality.

When I worked at an established tech company, no one was spending their days writing SAS code to do one-way ANOVAs. The data scientists were working on interesting, novel problems and could use whatever tools they wanted. R&D budgets were large so if you wanted the latest Mac + expensive commercial software, it was all good. Lots of Linux and a handful of Windows users as well.

Now at my manufacturing conglomerate, engineers use commercial software like Minitab or Design-Expert, while MS/PhD Statisticians use R or JMP. 

And FWIW, I highly suggest choosing the Stats route if you have the choice between a Stats PhD or Biostats PhD, even if that means having to take an extra semester or two of theory. You will be better off.

Edited by statsguy
Link to comment
Share on other sites

On 4/24/2021 at 11:05 AM, bayessays said:

Yeah, I got to the final stage of an internal transfer within a company to my dream location, and then the boss's boss wanted someone with a PhD.  I would definitely say just having a PhD versus a masters is a significantly bigger deal, in a variety of industries, than having a PhD from a high vs. low-ranked place.
 

Plenty of biostats PhDs go into data science, so you don't have to choose right away.  Really just think of a biostatistics PhD as the same as a statistics PhD with less options on the very theoretical/mathy end.  In fact, biostats PhDs often have a lot of good applied experience which is key for data science jobs, so I think a biostat PhD is a great option for that type of work!  But yes for something like genomics, you'll probably have to really specialize in the genetics end of things for your PhD.  But there are tons of data science jobs where a PhD in biostat would be helpful.

I would disagree with this based on my 10+ years of industry experience working at three fairly disjoint companies. Unless you're going into something hyper-specialized (e.g. genomics) a Stats PhD will generally be the better option.

Edited by statsguy
Link to comment
Share on other sites

2 minutes ago, statsguy said:

I would disagree with this based on my 10+ years of industry experience working at three fairly disjoint companies. Unless you're going into something hyper-specialized (e.g. genomics) a Stats PhD will generally be the better option.

Do you mean in the sense that employers might view a Stats PhD more favorably because they might believe the biostatisticians are too specialized or the stats PhDs are smarter?  I'm mostly basing this off me and lots of of PhD grads from same top 5 biostat program getting jobs at one of the big tech companies, so it seems like an easy way to break into that.   I think at the top biostat programs, the training is essentially identical to a top statistics program, so most people would end up with the exact same skills.  But obviously I'm only speaking from my limited perspective, so would be very interested to hear why this is the case based on your experience.

Link to comment
Share on other sites

22 minutes ago, bayessays said:

Do you mean in the sense that employers might view a Stats PhD more favorably because they might believe the biostatisticians are too specialized or the stats PhDs are smarter?  I'm mostly basing this off me and lots of of PhD grads from same top 5 biostat program getting jobs at one of the big tech companies, so it seems like an easy way to break into that.   I think at the top biostat programs, the training is essentially identical to a top statistics program, so most people would end up with the exact same skills.  But obviously I'm only speaking from my limited perspective, so would be very interested to hear why this is the case based on your experience.

On numerous occasions over the past 10 years I've seen the stereotype that Biostats = clinical trials + stats lite. It's unfortunate, and I personally don't agree with it, but it is what it is.  

10+ years ago when I graduated with a Stats PhD, the industry-bound Biostats PhDs at my university all went to companies like Baxter, Medtronic, Boston Scientific, Merck etc.   while the industry-bound Stats PhDs went to a very wide variety of companies - Pfizer, the IRS, Google, Amazon, and a startup (me). 

The coursework was also not the same, at least back then. Stats PhD had an extra year of theory, and the applied courses were much more general, with a slant towards machine learning, advanced regression, etc.  The Biostats courses were much more specialized - things like longitudinal data analysis, survival analysis, etc.  In fact, even categorical data analysis at that time was only offered in the Biostats department. 

Edited by statsguy
Link to comment
Share on other sites

41 minutes ago, statsguy said:

On numerous occasions over the past 10 years I've seen the stereotype that Biostats = clinical trials + stats lite. It's unfortunate, and I personally don't agree with it, but it is what it is. 

This is what I was trying to get at lol, industry has this exact clincal trial or validation monkey stereotype. And I am doing that right now and it sucks. This is why I said earlier biostat by the industry is less actual stats than data science hell even EE/CS now. Its more regulatory in nature. Writing focused, more businessy and law-ey. Honestly, I think some social science or public health majors with a decent handle on stats can succeed and enjoy in this area than someone who has been in STEM their whole life. 
 

1 hour ago, statsguy said:

You're overthinking this. How many positions have you had since you got your MS?

Just because you got stuck doing gruntwork in a biostat position doesn't mean industry is all SAS and one-sample t-tests.

When I was at a startup, our team worked around the clock to write algorithms to predict outcomes involving cancer diagnoses. Lots of R, Python, C/C++, etc. Based on what I later saw, most of the methods our team developed internally would blow away anything that was coming out of academia. But we were paid to help develop cancer screening methods that would ultimately save lives, not publish papers in journals whose audiences are 99% academics. Plus the whole thing about NDAs and confidentiality.

When I worked at an established tech company, no one was spending their days writing SAS code to do one-way ANOVAs. The data scientists were working on interesting, novel problems and could use whatever tools they wanted. R&D budgets were large so if you wanted the latest Mac + expensive commercial software, it was all good. Lots of Linux and a handful of Windows users as well.

Now at my manufacturing conglomerate, engineers use commercial software like Minitab or Design-Expert, while MS/PhD Statisticians use R or JMP. 

And FWIW, I highly suggest choosing the Stats route if you have the choice between a Stats PhD or Biostats PhD, even if that means having to take an extra semester or two of theory. You will be better off.

At a startup, things could be different yea and more innovative. 
 

It is my first job since I graduated MS a year ago, but in general just search on LinkedIn “Biostatistician” and compare the listings to Bioinformatics and Data Science even within Biotech. Biostat to industry = SAP, regulatory submissions, FDA/ICH guidelines, etc. The stats might be some power/sample size calculation. And then report a p value or confidence interval once the data comes in. This is like 1950s stuff. 
 

DS has its fair share of BS too like tableau and SQL but you can try to avoid those more easily without ruling like 90% out. Otherwise DS has signal processing/time series, ML/DL, multivariate analysis, causal inference, etc mentioned. Regulatory submissions are often not mentioned.

DS does more real actual statistics. I completely agree with your last point, though for me not having taken Real Analysis and having gotten a B+ in upper div lin alg as well as my MS math stat classes I feel a stat PhD will be tougher to get into. 

I currently am in a med-big company and it is quite product focused, getting the device out sort of thing. This involves reams of documentation. Analyses are mostly product validations and trials. It is high pressure during submission timeframes to get the writing correct down to the exact wording. Having talked to some people at even more well known biotech companies like Genentech, Biostat is not too much different there either.

Edited by untzkatz
Link to comment
Share on other sites

I think for biostatistics you just have to be more discerning when on the job market.

There are definitely a lot of advanced pharma/biotech jobs that do super cool things. But I am 100% certain that out of the jobs that require biostat PhD there are waaaaaaaaayy more jobs that are your "check the pvalue/regulatory mess" than your Apple Health research AI type stuff. The  sinecure  statistics jobs are probably less prevalent for standard PhD statistics positions because getting a Statistically trained PhD to perform simple methods is only profitable and necessary in pharma/biotech.

There's just no other industry where millions of R&D and next-gen science research hinges on the reputation of  a crack PhD team sending a jaded bureaucrat who hasn't read a statistic paper since 1995,  a sas7bdat  file that computes a ANOVA table.

That same effect doesn't exist for Biostat/Stat PhDs in tech/finance (in finance regulatory/compliance needs more finance background than Stats).

Therefore it becomes way more important to say no to jobs as a bio-statistician. If Facebook/Citadel offers you a statistician position, it's probably going to be okay and fairly interesting, but if Pfizer does the same thing that could range from doing crazy causal inference research to formatting the shadows of a pie chart on a report.  I interviewed for some biostats type jobs out of undergrad, and it's very clear from the interview what type of job it's going to be. So you just have to say no when the interview is clearly just a joke.

I also suspect that all the cool jobs probably go to prestigious biostat programs and if you're from a school outside of Harvard, Michigan, Berkley, Upenn, JHU, UNC it'll be very difficult to get your foot into the door to top bio-statistician jobs.

FWIW, off a gut feeling I feel like the impact of a Biostats vs Stats would matter more as prestige of program goes down. E.g Harvard Biostats vs Harvard Stats probably doesn't make a different in industry. Duke Stats vs Duke Biostats probably matters quite a bit more. Rutgers Stats vs Rutgers Biostats is probably huge difference.

 

 

 

 

Edited by trynagetby
Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...

Important Information

This website uses cookies to ensure you get the best experience on our website. See our Privacy Policy and Terms of Use