Bias Adjustment: Local Control Analysis of Radon and OzoneThursday, September 29, 2016 - 2:45pm
Thursday, September 29, 2016 - 2:45pm
Where: LeConte College, Room 210
Speaker: Stan Young
Affiliation: National Institute of Statistical Sciences
Large, complex observational data sets typically present research opportunities, but also problems that can lead to false claims. In Big Data, the standard error of an average effect estimate goes to zero as sample size increases, so even small biases can lead to declared (but false) claims. The average of a treatment difference, a so called main effect, can often be meaningless when there are interactions with confounders that create local variation in effect-sizes. Scientists need statistical methods that can deal simply and efficiently with these sources of bias. Here, we demonstrate use of a new statistical analysis strategy, Local Control. The basic concept for Local Control is to first cluster the data and then do analysis within the clusters. Our first case study illustrates reduction of bias in an environmental epidemiology data set, radon. Our second study uses Local Control on a time series air quality example. By detecting interactions, scientists can produce more realistic and more relevant analyses that reduce the bias typically implied by the variety and heterogeneity of Big Data. Our results are rather surprising. Radon appears to protect against lung cancer, not cause it and ozone appears to have little or no effect on acute mortality, rather than cause deaths.
Announcement—Stan Young, Bias Adjustment: Local Control Analysis of Radon and Ozone
Presentation—Bias Adjustment: Radon and Ozone. Young, Obenchain, & Krstić. V02
Fair Treatment Comparisons in Observational Research. Lopiano, Obenchain, & Young, 2014.
Advancing Statistical Thinking in Observational Health Care Research. Obenchain & Young, 2013.