Throughout my career in the medical device industry, I’ve often found that engineers have a flabbergasting aversion to statistical analysis. And it’s crippling: I hardly see engineers during clinical trials, the final essential step in the medical device development process. To engineers, the clinicians seem to have the statistical know how and leave the numbers to them. Their conclusions are rarely challenged.
Why would so many professionals cut themselves out of a product at the pointy end of development, based on mathematics?
There’s a language around statistics that engineers find foreign. Engineers typically take one subject in statistics during their university programs, but it usually comes too early in their studies and without the right real-world context. It’s long forgotten by the time they practise as engineers, by which time the sight of a confidence interval seems an alien code.
Withdrawing from the subject, however, is greatly disempowering. It is yet another reason that engineers struggle for visibility in their organisations. The maths required, in all honesty, is a trifle compared to what engineers are capable of. What is at issue, then, is their openness to a different way of thinking.
In the last few years however, there’s a rebranding taking place that could make statistics more alluring for engineers: data science. Statisticians loathe the term, but many begrudgingly concede that the big data era requires broader skills in statistical analysis than plain mathematics.
A recent book on data science is Doing Data Science: Straight Talk from the Frontline by Cathy O’Neil and Rachel Schutt of Columbia University (O’Reilly Media). Rachel, a statistician, surveyed her data science class on their skill sets, and found a combination of statistics, mathematics, computer science, machine learning, communication, data visualisation and domain expertise. The authors quote Metamarkets CEO Mike Driscoll on the subject: “Data science is the civil engineering of data.” Interesting.
Increasingly, statisticians acknowledge that most statistical analyses are not performed by statisticians (for more, see Jeff Leek’s post). Engineers should have the breadth of skill, from mathematics to computing to domain expertise to communication (we are Expressive Engineering after all), to make outstanding statisticians. The willingness to embrace statistics will greatly advance their professional cause in the 21st century.
