CMStatistics 2015: Start Registration
View Submission - CMStatistics
B1354
Topic: Contributions on variable selection Title: Statistical analysis of data from restricted randomised experiments via shrinkage methods Authors:  Kalliopi Mylona - University of Southampton (UK)
Steven Gilmour - University of Southampton (UK)
Sadiah Aljeddani - University of Southampton (United Kingdom) [presenting]
Abstract: For model selection purposes in experimental contexts, researchers often use stepwise regression or subset selection. In situations involving restricted randomization, such as block experiments and split-plot experiments, this has to be done manually and, especially for experiments involving multiple responses, it involves numerous model estimations. We propose a modification of penalized least squares, for the analysis of data from experiments conducted under restricted randomization, which performs model selection and model estimation simultaneously. Under the penalized generalized least squares, non-significant variables are more likely to be dropped out the models. We demonstrate the usefulness of the approach using various practical examples, and study its properties in a simulation study.