Speaker：Prof. Wei-Yin Loh (University of Wisconsin, Madison)
Topic：A Decision Tree Approach to Fitting Models to Missing Data
Speaker：Prof. Wei-Yin Loh (University of Wisconsin, Madison)
Time：May 29 (Friday) , 2026, 10:40-11:30
Place： 4F-427, Assembly Building I
Abstract
An old and still difficult question in statistics is how to fit regression models to data with missing values in the predictor variables. "Regression" is considered generally here, ranging from classical linear and logistic regression to modern machine learning methods. For more than fifty years, a standard practice has been imputation, which often requires assumptions that are often neither theoretically justifiable nor practically verifiable. Missingness itself, however, can contain useful information that imputation destroys. We show how the GUIDE classification and regression tree algorithm can build prediction models without requiring imputation. The GUIDE software and background material are available from https://pages.stat.wisc.edu/~loh/guide.html.