Modernizing Mixed Model Prediction

Modernizing Mixed Model Prediction

Sunil Rao, University of Minnesota, Biostatistics & Health Data Science

Abstract

Mixed models have widespread appeal in many areas of statistical modeling from biostatistics to small area estimation.  I will conduct a tour of some recent developments in both methodology and application related to mixed model prediction (MMP) that I have contributed to or are outgrowths of those methods.  The talk will review earlier work including observed best prediction (OBP) and classified mixed model prediction (CMMP) and then discuss some newer developments.  This includes multivariate MMP for prediction of the genetically controlled portion of DNA methylation in cervical cancer and contextual vulnerability analysis for predicting overdose deaths.  Finally if time permits, I will discuss MMP in the context of making projections potentially outside the range of the training data.   This is joint work with a number of  PhD students over the years, Jiming Jiang at UC, Davis and Thuan Nguyen at Oregon Health and Science University.

Bio

J. Sunil Rao is Professor of Biostatistics in the SPH at the University of Minnesota.  He is also the Director of Biostatistics at the Masonic Cancer Center.  He is also Professor Emeritus and Founding Director of the Division of Biostatistics at the University of Miami.  His research interests include mixed model prediction and selection, high dimensional modeling, small area estimation, information theory and machine learning.  He also has research interests in health disparity problems with a focus on lung cancer outcomes.