Re: [R] logistic regression: differential importance of regressors

From: Frank E Harrell Jr <f.harrell_at_vanderbilt.edu>
Date: Fri 20 May 2005 - 07:22:30 EST

Greg Trafton wrote:
> Hi, All. I have a logistic regression model that I have run. The
> question came up: which of these regressors is more important than
> another?
>
> (I'm using Design)
>
> Logistic Regression Model
>
> lrm(formula = iconicgesture ~ ST + SSP + magnitude + Condition +
> Expertise, data = d)
>
> Coef S.E. Wald Z P
> Intercept -3.2688 0.2854 -11.45 0.0000
> ST 2.0871 0.2730 7.64 0.0000
> SSP 0.7454 0.3031 2.46 0.0139
> magnitude -0.9905 0.6284 -1.58 0.1150
> Condition 0.9506 0.2932 3.24 0.0012
> Expertise 0.8508 0.2654 3.21 0.0013
>
> The real question is that, since both ST and SSP load significantly
> into the model, how do I show that ST has a bigger/smaller/similar
> effect than SSP?
>
> thanks in advance!
> greg
>

One thing you can do is to compute what proportion of the total likelihood ratio chi-square statistic is due to each variable by removing one at a time and looking at difference in Model L.R. (assuming both have same observations missing). Note that you are making heavy linearity assumptions.

You can also use the bootstrap to get a confidence interval on the rank of the chi-square statistic a variable has among all competing chi-squares.

Frank

-- 
Frank E Harrell Jr   Professor and Chair           School of Medicine
                      Department of Biostatistics   Vanderbilt University

______________________________________________
R-help@stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Received on Fri May 20 07:28:52 2005

This archive was generated by hypermail 2.1.8 : Fri 03 Mar 2006 - 03:31:56 EST