Standard model assessment techniques such as residual plots or Akaike's information criterion can be difficult to use or provide limited insight into model fit when applied in non-standard regression contexts such as random effects or mixture models. This paper considers the application of Bayesian ideas to model choice for a large consumer preference dataset where non-standard models appear to be needed but it is unclear which model is most appropriate. In addition, it is shown that suitably chosen graphical methods can provide insights into which of a set of competing models is most useful for different subsets of the data.
Back to Research page.
Send me e-mail at ipardoe at lcbmail.uoregon.edu
Last updated: September 9, 2003
The views and opinions expressed in this page are strictly those of the page author. The contents of this page have not been reviewed or approved by the University of Oregon.
© 2003, Iain Pardoe, Lundquist College of Business, University of Oregon