Risk and uncertainty

Failure To Communicate: Why big data analytics projects fail and how to save them

If you are like most practitioners of analytics, you went into the field because you are good with numbers, and you want to help people improve the quality of decisions on important issues. But many of us spend too much time struggling to organize poorly structured data and debugging complex spreadsheets or code and too little time engaging directly with our clients to help them clarify their objectives and decisions, brainstorm better decision options, and explore, visualize and understand the results. Without this kind of interaction, the analysis may fail to address the issues they really care about. Even if it does, clients may not develop the confidence to rely on the insights, and you end up frustrated that your hard work fails to be properly appreciated.
Read the article at Orms Today Magazine, Dec 2019

Testing hypotheses about causation

In 2002, I developed a statistical framework for testing whether your data provides statistically significant support for the hypothesis that A causes B. I published only one conference paper with some colleagues on the idea before moving on to other things, but I’ve always felt the idea was pretty novel and interesting.

Scroll to Top