Dear Ahmet,Dear Mr. Fox
Sorry for the slow response, but I've been busy all today, coincidentally teaching a workshop on logistic regression.
Tom Blackwell sent you a useful suggestion for interpreting coefficients on the odds scale. If you want to trace out the partial relationship of the fitted probability of response to a particular predictor holding others constant, you can set the other predictors to typical values and let the predictor in question vary over its range, transforming the fitted log-odds to the probability scale.
You may be interested in my effects package (on CRAN or at <http://socserv.socsci.mcmaster.ca/jfox/Misc/effects/index.html>), which makes these kinds of displays for linear and generalized-linear models, including those with interactions.
Regards, John
At 03:06 PM 6/3/2003 +0300, orkun wrote:
John Fox wrote:
At 11:54 AM 6/3/2003 +0300, orkun wrote:thank you
in logistic regression,
I want to know that it is possible to get probability values of each predictors by
using following formula for each predictor one by one (keeping constant the others)
<<< exp(coef)/(1+exp(coef)) >>>
Dear Ahmet,
This will almost surely give you nonsense, since it produces a fitted probability ignoring the constant in the model (assuming that there is one), setting other predictors to 0 and the predictor in question to 1. What is it that you want to do?
I hope that this helps, John
Say, I just want to find each predictor's particular effect on dependent variables.
Actual model is to prepare landslide susceptibility map on GIS. So I want to know
what the effect as probability value comes from each predictor. For instane what is the effect
of slope on landslide susceptibility. Should I keep others constant ?
kind regards
----------------------------------------------------- John Fox Department of Sociology McMaster University Hamilton, Ontario, Canada L8S 4M4 email: [EMAIL PROTECTED] phone: 905-525-9140x23604 web: www.socsci.mcmaster.ca/jfox -----------------------------------------------------
thank you very much all.
Because of related to your answer. I ask you directly if you don't mind
I studied several ways after my email.
I wonder whether pgeo<-predict.glm(glm.ob,type="terms")
gives same result with probability value I asked before.
I tried on it. But it gives "Error in rep(1/n,n) %*% model.matrix(object): non conformable
arguments" .
By the way , your teaching notes is available on the internet ?
cordially
can y
______________________________________
______________________________________
The views and opinions expressed in this e-mail message are the ... {{dropped}}______________________________________________ [EMAIL PROTECTED] mailing list https://www.stat.math.ethz.ch/mailman/listinfo/r-help
