[R] binomial GLM quasi separation

lincoln miseno77 at hotmail.com
Fri Oct 14 11:32:24 CEST 2011


As you suggested I had a further look at the profile by changing default
values of stepsize (I tried to modify the others but apparently there was
any change).
Here they go the scripts I have used:

> dati<-read.table("simone.txt",header=T,sep="\t",as.is=T)
> glm.sat<-glm(sex~twp+hwp+hcp+hnp,binomial,data=dati)
Mensajes de aviso perdidos
glm.fit: fitted probabilities numerically 0 or 1 occurred 
>pp<-profileModel(glm.sat,quantile=qchisq(0.95,1),objective="ordinaryDeviance",which=4)
Preliminary iteration . Done

Profiling for parameter hcp ... Done
>pp1<-profileModel(glm.sat,quantile=qchisq(0.95,1),objective="ordinaryDeviance",which=4,stepsize=20)
Preliminary iteration . Done

Profiling for parameter hcp ... Done
>pp2<-profileModel(glm.sat,quantile=qchisq(0.95,1),objective="ordinaryDeviance",which=4,stepsize=100)
Preliminary iteration . Done

Profiling for parameter hcp ... Done
>plot(pp)
>mtext("Default stepsize",adj=0,cex=2,line=1)
>dev.new()
>plot(pp)
>mtext("Stepsize=20",adj=0,cex=2,line=1)
>dev.new()
>plot(pp)
>mtext("Stepsize=100",adj=0,cex=2,line=1)

And these are the plots as they look like:
http://r.789695.n4.nabble.com/file/n3904261/plot1.png 
http://r.789695.n4.nabble.com/file/n3904261/plot2.png 
http://r.789695.n4.nabble.com/file/n3904261/plot3.png 

I have tried to understand what is going on but I don't know how to
interpret this.
It's quite a long time that I am trying to solve this but I have not been
able to. Here (  http://r.789695.n4.nabble.com/file/n3904261/simone.txt
simone.txt  ) I attach a subset of the data I am working with that comprises
the variables specified in the above glm model and by the way the "funky"
variable called "hcp".
Thank you for take your time to help me in this.

--
View this message in context: http://r.789695.n4.nabble.com/binomial-GLM-quasi-separation-tp3901687p3904261.html
Sent from the R help mailing list archive at Nabble.com.



More information about the R-help mailing list