[R] linear models and colinear variables...

Peter Gaffney petertgaffney at yahoo.com
Fri Jul 2 02:53:43 CEST 2004


> When you do this, you are including all the
> interaction terms.
> The * indicates an interaction, as opposed to +.

In this particular case I need to do exactly this;
this is a study of antibiotic resistance - two of the
variables respectively are type of bacteria and
antibacterial agent. The evolutionary/epidemiological
behavior of each pairing of these factors is
different.  Can I remove some lower order terms; for
example, if I get rid of Bugtype:Usage.level.ofdrug
and Drugtype:Usage.level.of.drug will 
Bugtype:Drugtype:Usage.level.of.drug still be valid?

> If you select predictors on the basis of which ones
> are
> significant, then the final significance levels
> don't mean much,
> usually.  Remember, 1 out of 20 will be significant
> at .05 even
> if you are using random numbers.

This is an excellent point; were I to proceed I would
need to select based strictly on removing from
collinear pairs or groups of explanatory variables,
probably according to an a priori established ordering
of classes of variables; ie B:D:U might be more
interesting than B:U or D:U or B:D:U:ICU, so remove
collinear variables from the latter three first,
irrespective of statistical significance.

Thanks for you help. :-)


More information about the R-help mailing list