How can I be a biostatistician freelancer?

It’s not easy, however it’s doable and fully depends on your attitude, ability and willingness to learn. Let me enumerate some important things:

  1. good biostatistician must expose the ability to learn medical, pharmacological, and diagnostics-related terms and processes. Without that, one is going to be just a statistician doing some calculations in medical research, but without the knowledge of what he is doing, why this is important, what might be important to researchers, how to better check the data, what to propose when asked for help by researchers. In other words – lack of this ability makes the biostatistician almost blind during the analysis.

    In clinical research, writing a synopsis for a project, statistical input for the protocol or SAP (statistical analysis plan) is a perfect example of the situation, where this ability is extremely helpful, just priceless. Yes, sponsors should come with the ready set of hypotheses to be statistically verified, but this is only in theory. It’s more than likely biostatistician will be asked many times for advice on how to conduct a certain kind of analysis.

    How is one going to respond well having even no rough idea about the subject of a trial and how everything works? Statistics in medical research is not very advanced (relatively), but it requires strong understanding of what is going on, how to interpret results.

    This is the place where outliers (lying far from the majority of observations) might be perfectly OK from clinical perspective and observations seem perfectly OK might convey worrying news.

    This is visible much better in the exploratory analysis (in evidence-based medicine), when biostatistician helps a researcher in writing a thesis or article or in conducting scientific research. They make a closely coupled team entering an unknown area. No ready hypotheses (just some anticipation and initial thoughts), mostly ad’hoc ideas, unexpected situations, strange results of calculations (indicating errors or a new discovery) and so on.

    It doesn’t mean biostatistician should own PhD in medicine! Not at all. But he should quickly learn new things from the domain of medicine, diagnostics or pharmacy, acquire it and accumulate for further use.

    Can one be a good biostatistician without this ability? My constant answer is: no, and don’t even try to start, just save researcher’s souls and don’t bother them, really. Sorry for sounding harshly, but that’s the simple truth.

    If you feel you like medicine, go ahead, it’s your chance.

  2. It will be really hard to find an opportunity to start without a clear path of education or past experience. Not many people/companies want to risk and possibly lose their time=money by trusting someone who just starts.

    But it doesn’t mean it’s impossible. I did it. Start your journey from searching someone you could help, support in his/her research. Don’t ask for money. The *opportunity* you get is worth much more than any money – it builds your experience and portfolio. If the process is successful, you can humbly ask for written recommendation. Perhaps the researcher, full of gratitude, will list you as a co-author? That’s priceless.

    Then just repeat the step above. Find another researcher and another one and so on. Build your professional network. Ask for being recommended. Ask your contacts if someone searches for a similar help.

  3. When your portfolio is ready, try a level up. Find a company willing to give you another chance and offering you at least part-time job. Repeat this once or twice. Try to stay for a longer time, don’t jump from a project to project too fast, learn the specific of the industry.
  4. Learn both R and SAS. R is very common in evidence-based medicine, SAS still (and probably “forever and one day”) strongly dominates in clinical research.

    Personally, I didn’t learn SAS, but this my conscious decision resulting in limiting my own chances. It’s up to you, but I recommend learning it. SAS for students and learning is free: Free Statistical Software, SAS University Edition

  5. Read books about biostatistical methods, testing for bioequivalence /superiority / non-inferiority, TOST, pharmacokinetics and -dynamic analysis,survival analysis, sample size calculation (not only for simple t-test or proportion test), linear and nonlinear modeling, mixed effects models, GEE, missing data imputation algorithms (not only the simplest ones like LOCF/BOCF, but also modern methods like MI, MICE), bootstrapping. Learn about dealing with violations of statistical methods, because “small data” (very small data sets) are common in medicine.

    Forget about modern data science fancy-schmancy methods, you will play with 20–50–100–1000 (depending on the project, phase of a clinical trial) most of the time. And use classical statistical inference. You must understand it well.

    Just a few examples of books:

    1. Design and Analysis of Clinical Trials: Concepts and Methodologies, 3rd Edition (read it, really!)
    2. Clinical Trial Data Analysis Using R
    3. Handbook of Parametric and Nonparametric Statistical Procedures, Fifth Edition (9781439858011): David J. Sheskin: Books (must have, trust me)
    4. Design and Analysis: A Researcher’s Handbook (4th Edition) (9780135159415): Geoffrey Keppel Professor Emeritus, Thomas D. Wickens: Books
    5. Biostatistical Design and Analysis Using R: A Practical Guide

I hope this helps a bit

Leave a Comment

Your email address will not be published. Required fields are marked *

/* add by OCEANUS */