Should I pursue a PhD in Statistics?

Question

I am trying to decide whether or not I should take a PhD program in Statistics. I am not inclined to get a position at a university after my PhD, my goal is rather to get hired by Contract Research Organizations involved in Clinical Trials, either as a Biostatistician or Statistical Programmer. At the moment, I like working with SAS so I am more inclined to become a Statistical Programmer.

Would the Clinical Trials companies hire a PhD holder for their Statistical Programmer position? From what I see on the job postings, companies typically asks for Master's degree in Statistics (or sometimes even just a Bachelor's degree) for Statistical Programmer positions, but I just wanted to hear opinions of people who actually works in Clinical Trials for private companies.

Thank you,

Answer 1

I have a PhD in statistics, where I specialised in Bayesian theory. These days I am doing work as lead statistician on RCT research in a couple of health projects. This involves RCT planning and execution, analysis of data, and reporting outcomes of trials. On the basis of my experience, this is what I think:

• For the vast majority of work doing statistical programming in health, a Masters-level education in statistics would be sufficient. The main skills you will need for practice are good skills in using statistical computing (e.g., SAS, R, etc.) to organise, clean, and analyse data, and create routines for automated reproducible analysis. Most of the models you use are likely to be commonly used model forms that have already been programmed into statistical software (e.g., regression, GLMs, GLMMs), and it is rare that you need custom models. You should put in the time to understand these basic model categories deeply, and learn the implementation of these models in statistical programming.

• A PhD is likely give you a deeper knowledge of theory than most others in your field who lack that background, and better mathematical skills. It also gives you practice at the process of research leading to peer-reviewed published work. This level of study gives you very solid "first principles" knowledge of statistical theory and models, which gives you an advantage when you encounter problems requiring some variation of standard models or custom models.

• Aside from general theory knowledge, and improved mathematical ability, the value of a PhD depends a great deal on the relevance of your research topic to your future field. If you undertake a research project in the field of statistical programming for RCTs, that will be very helpful for a future career in that field. If your topic is not relevant to your future field (as in my case) the value you will obtain will just be a general improvement in your theory and mathematical abilities, and broader knowledge of statistics.

• Even with an irrelevant project, a PhD in statistics is going to give you some training that is useful in a general sense (better theory knowledge, better maths, etc.). Although there is value in this program, there is also a big opportunity cost. If you spend a standard full-time period of four years doing a PhD, that is going to be at the expense of four less years of professional experience in the industry. Using the equivalent amount of time practicing in the field is likely to give you much more skill in the day-to-day operations in that field than undertaking a PhD.

• If you're even moderately undecided about a PhD, I suggest you don't do it. A PhD candidature is a major commitment requiring a hard slog through a lot of road-blocks. It is rarely smooth, and the academic landscape is littered with the bodies of PhD drop-outs. For what you want to do, I would suggest trying to get industry experience as early as possible, and then consider later postgraduate education once you have a bit of experience.