Data science is all about working large data sets. Data scientists often perform experiments, in order to find interesting and useful correlations. But it’s easy to perform analyses that aren’t accurate, or that help you to draw conclusions that are less than accurate. How can you be sure that your analysis is robust? Use techniques that embrace your doubts, and allow you to demonstrate to yourself and others that the correlations you’re seeing really exist. In this talk, Hillary Green-Lerman introduces this problem, and then describes solutions which allow you to confidently describe the conclusions you’re drawing from your data-science experiment.