At the second annual Women in Data Science conference, Stanford professor of statistics Susan Holmes explored some of the challenges she faces and the methods she uses in her research on the human microbiome.
“Even if you don’t know anything about it, you all have a microbiome that you should care about,” she said. It is a factory that makes the chemicals that we need for our brain, for our body to function.”
Holmes called for biomedical researchers to take steps to ensure their data and methods are open source and to promote the reproducibility of their work. That, she said, will help overcome worries about whether or not you’re making the right call in the face of a multiplicity of choices and analyses.
“We can make our work reproducible and not spend our time oscillating about what’s right and what’s not right.” Holmes also urged younger researchers in particular to embrace messy data sets and ambiguous results. “Real problems, and I think you have to be honest about them, are quite heterogeneous and messy. If you have a really clean story, it’s probably wrong, because it would have been found before.”