On Tue, 23 Sep 2003 12:42:08 +0200, Frank E Harrell Jr wrote: > On Fri, 19 Sep 2003 13:49:09 GMT > Marc Schwartz <[EMAIL PROTECTED]> wrote:
SNIP >> Some of the questions you need to ask: >> >> 1. Do the 'non-significant' (NS) variables contribute to improving the >> model fit? Compare model fit metrics with and without the NS variables. >> >> 2. Do the NS variables contribute to the interpretation of >> the model for the target audience/users or detract from it? > > Marc - these are stickier issues than you implied. Frank, No disagreement. As I was responding a plethora of thoughts were going through my mind and I put them down quickly. Perhaps too quickly and too superficially. To your point, there are many caveats surrounding these issues and each requires more indepth consideration than what is likely possible in this forum. Hence I referenced your book, which it sounds like the gentleman has available to him... :-) >> 3. If you remove the NS variables from the model, are there other >> variables in your dataset that might be considered? Do the 15 constitute >> all of your available data or only a subset? >> >> 4. Would the inclusion of the NS variables result in over-fitting of the >> model? > > Removal of NS variables does not help with overfitting. It just covers it up. Quite true. > You made some good points. -Frank Thanks for your comments and clarifications as well. Regards, Marc . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
