s-news
[Top] [All Lists]

Proposing models

To: <s-news@lists.biostat.wustl.edu>
Subject: Proposing models
From: "Roger Gill" <R.C.GILL@soton.ac.uk>
Date: Thu, 27 Jan 2005 14:24:40 -0000
Dear All,

I would like to do the following.

Given factors x1,x2,......,xn I define a model in terms of its generators.
For example y~x1*x2*x3+x4+....+xn.  Given this model I would like to move to
a neighbouring model ensuring that any model is hierarchical.  So for the
given model I could add any two way interaction not already in the model, or
choose to remove the three way interaction from the current model.  Clearly
I would not want to remove any two way interaction if the three way is
present.

I would initially like to code this in R, eventually moving to C, and hence
would like some inspiration as to how I could achieve my goal.

Any ideas would be much appreciated

Best Wishes

Roger Gill


<Prev in Thread] Current Thread [Next in Thread>