There is an epidemic of statistical dunderheads working in the biotech industry. This epidemic probably plagues other sectors of the economy as well, but I'm not qualified to speak to that.
The reason for this complete lack of statistical knowledge (I think) is that statistics is not a part of the standard engineering curriculum. You get differential and integral calculus like crazy, but just one semester of basic engineering statistics and here's your diploma.
And as with most of undergraduate academia, it's not practical.
At my first job, we used the statistical software program - JMP - a lot. We were making a minimally invasive glucose monitor called the, GlucoWatch® Biographer and my entire job as a research engineer was to run in-house clinical studies and correlate the biographer performance against over-the-counter glucose meters. So we did a lot of linear correlations, I got to understand what p-values meant. And I figured out the primary purpose of engineering a system was to figure out what was signal and what was noise.
I think I might have even landed my second job because I knew how to use JMP. In fact, my second week on the job, the boss had his entire group go get JMP training in San Francisco where I had the luck of sharing a computer terminal with him.
Whatever the case, understanding enough statistics to know what tests are applicable when is really important. And when your group gets big enough that sending your team to off-site training becomes impractical, there is Dr. Thomas Little who will send practical stats gurus to train you.
Dr. Tom trained us (in a computer room) setting and a lot of this stuff was new at the time I learned it. ANOVA... Multivariate Analysis. Why use backwards stepwise regression... how to read the normal quantile plot... Capability... Control Charting.... All the things that are relevant to monitoring a production campaign.
When you get out of the class, you've leveled up in the world of biologics manufacturing and you look around and wonder why maintaining spreadsheets of cell culture data qualifies as plant support. You also start wondering why process development spends more time swinging male genitalia over higher titers rather than defining critical process parameters (CPPs) and identifying proven acceptable ranges (PARs).
Dr. Tom is pretty well-known in the world of biologics. I run into his team of consultants every third place I go. If your team isn't making statistically-sound, data-drive decisions, you seriously need to give him a call.
Call Dr. Tom
@zymergi
The reason for this complete lack of statistical knowledge (I think) is that statistics is not a part of the standard engineering curriculum. You get differential and integral calculus like crazy, but just one semester of basic engineering statistics and here's your diploma.
And as with most of undergraduate academia, it's not practical.
At my first job, we used the statistical software program - JMP - a lot. We were making a minimally invasive glucose monitor called the, GlucoWatch® Biographer and my entire job as a research engineer was to run in-house clinical studies and correlate the biographer performance against over-the-counter glucose meters. So we did a lot of linear correlations, I got to understand what p-values meant. And I figured out the primary purpose of engineering a system was to figure out what was signal and what was noise.
I think I might have even landed my second job because I knew how to use JMP. In fact, my second week on the job, the boss had his entire group go get JMP training in San Francisco where I had the luck of sharing a computer terminal with him.
Whatever the case, understanding enough statistics to know what tests are applicable when is really important. And when your group gets big enough that sending your team to off-site training becomes impractical, there is Dr. Thomas Little who will send practical stats gurus to train you.
Dr. Tom trained us (in a computer room) setting and a lot of this stuff was new at the time I learned it. ANOVA... Multivariate Analysis. Why use backwards stepwise regression... how to read the normal quantile plot... Capability... Control Charting.... All the things that are relevant to monitoring a production campaign.
When you get out of the class, you've leveled up in the world of biologics manufacturing and you look around and wonder why maintaining spreadsheets of cell culture data qualifies as plant support. You also start wondering why process development spends more time swinging male genitalia over higher titers rather than defining critical process parameters (CPPs) and identifying proven acceptable ranges (PARs).
Dr. Tom is pretty well-known in the world of biologics. I run into his team of consultants every third place I go. If your team isn't making statistically-sound, data-drive decisions, you seriously need to give him a call.
Call Dr. Tom
@zymergi
No comments:
Post a Comment