Message-ID: <429711CB.8020301@pdf.com>
Date: 2005-05-27T12:25:47Z
From: Spencer Graves
Subject: Power set
In-Reply-To: <4297064A.1040103@vanderbilt.edu>
Hi, Laura:
Have you considered "regsubsets" in library(leaps)? Also, have you
done an R site search "www.r-project.org" -> search -> "R site search"
for something like "all subsets regression"?
Consistent with Frank Harrell's comment, comments, I suggest you
randomly permute your response variable a hundred or a thousand times
and count how many times the procedure finds a non-null model. This
might help you calibrate the procedure. The archives include an email
from Frank Harrell citing two relevant articles.
Hope this helps.
spencer graves
Frank E Harrell Jr wrote:
> Laura Holt wrote:
>
>> Hi again!
>>
>> I have a data.frame with the columns y, x1, x2, x3.
>>
>> I would like to fit linear models with one variable at a time,
>> then 2 variables at a time, and then 3.
>>
>> Makes me think of a power set.
>
>
> Makes me think of irreproducible results if you use the output to select
> a single model :-)
>
> Frank Harrell
>
>>
>> Anyhow, is there a function to produce the right hand side of the
>> formulas, please?
>>
>> thanks,
>> Laura Holt
>> mailto: lauraholt_983 at hotmail.com
>> R Version 2.1.0 Windows
>
>