Title:

Variable Selection or Variable Assessment?

Speaker:
Drs. R. D. Meyer & R. G. Wilkinson Lubrizol Corporation
Time:

Friday, September 6,1996 3:30 pm - refreshments
4:00 pm - talk

Location:
Room 327, Yost Hall

Abstract

Variable-selection regression methods are oriented towards selecting a single model as the vehiole for further inferences. The appropriate inference about variables not included in the model is unclear; the conclusion that they have no effect may be misleading. In many situations, the objective of the statistical method should be to assess the relative importance of every variable. The term vadable assessment is more descriptive of this objective. A method for variable assessment that makes use of the Bayesian model-selection methodology will be described. The marginal posterior probability that a variable is needed in the model is a measure of its importance. Using the Gibbs sampler for computation greatly reduces CPU requirements and also allows for extension of the model to one that allows for outliers. A simulation study demonstrates that the method has good statistical properties.