"Improved Prediction and Flagging of Extreme Random Effects for Non-Gaussian Outcomes Using Weighted Methods

Speaker: John Neuhaus, PhD, MS, MA, Professor of Biostatistics, UCSF

In this talk, I will present some recent work on methods to predict and identify or flag extreme random effects for non-Gaussian outcomes that more heavily weight contributions from extreme random effects. I will show that these new weighted methods can provide more accurate prediction and higher correct flagging rates than standard best prediction methods, while controlling the incorrect flagging rates. I will illustrate the results with data from a study of hospital readmissions among children with asthma.