Statistics Seminar - Jerry F. Lawless - Bias in Statistics: What is it, How Does it Arise, and Can we Avoid it?
Speaker: Jerry F. Lawless (University of Waterloo)
Title: Bias in Statistics: What is it, How Does it Arise, and Can we Avoid it?
Abstract: The term bias has numerous meanings in statistics and, more broadly, in science. A seemingly narrow statistical definition refers to the tendency of an estimation procedure to over- or under-estimate a target parameter in some population. However, when we think about bias in scientific inference more generally, we find that this definition, and an extended version that also addresses testing of hypotheses, covers a great many settings. What is complex and subtle, though, are the many ways that bias can arise and the steps we might take to avoid or reduce it. In this talk I will describe some of the ways that bias arises in surveys or studies in areas such as health research, economics and social science, and some ways to mitigate bias. I will also consider some important issues related to data science, involving big data, data integration and algorithms for classification and prediction.