The Effects of Different Missing Mechanisms on Stepwise Variable Selection
Objective To study the effects of different missing mechanisms on stepwise variable selection in linear regression. Methods Data with different missing mechanisms and patterns were simulated and the performance of stepwise variable selection was evaluated using: the number of authentic variables that were selected and loss function I. E.. Results Stepwise selection using the complete data had the best performance, and the missing pattern had a greater impact on than missing mechanism type. Conclusion When performing stepwise selection on data with missing values, it is important to understand the type of missing pattern and not just the missing mechanism.
NETL
NSTL
万方数据
维普
