R help - Aug 2004 - logistic regression

I have a system with a binary response variable that was hypothesized to 
follow a simple logistic function. The relationship between the continuous 
independent variable and the logit is clearly not monotonic. I have two 
questions. 1) Can anyone recommend a reference that describes my modeling 
options in this case, and 2) what facilities does R have to deal with this 
situation?

Thanks,
Kevin Cummins
School of Medicine
University of California San Diego

On Mon, 2 Aug 2004, Kevin Cummins wrote:
> I have a system with a binary response variable that was hypothesized to
> follow a simple logistic function. The relationship between the continuous
> independent variable and the logit is clearly not monotonic. I have two
The logit is a monotonic function...or do you mean something different
here?
> questions. 1) Can anyone recommend a reference that describes my modeling
> options in this case, and 2) what facilities does R have to deal with this
> situation?
Look at glm(), its man page and the references therein.

hth,
Z
> Thanks,
> Kevin Cummins
> School of Medicine
> University of California San Diego
>
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://www.stat.math.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide!
http://www.R-project.org/posting-guide.html
>

My suggestion would be a generalized additive model, which you can fit with
gam() in the `mgcv' package.  If you are adventureous you can also try it
`by hand' using glm() with spline terms.  You might also try lrm() with
splines (in the `Design' package) as described in Prof. Harrell's book
`Regression Modelling Strategies'.

HTH,
Andy
> From: Kevin Cummins
> 
> I have a system with a binary response variable that was 
> hypothesized to 
> follow a simple logistic function. The relationship between 
> the continuous 
> independent variable and the logit is clearly not monotonic. 
> I have two 
> questions. 1) Can anyone recommend a reference that describes 
> my modeling 
> options in this case, and 2) what facilities does R have to 
> deal with this 
> situation?
> 
> Thanks,
> Kevin Cummins
> School of Medicine
> University of California San Diego
> 
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://www.stat.math.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide! 
> http://www.R-project.org/posting-guide.html
> 
>

Have you considered function "glm" with some polynomial in the 
independent variable?  Regarding a more general reference, have you seen 
Venables and Ripley (2002) Modern Applied Statistics with S, 4th ed. 
(Springer)?  If the answer to either of these questions is "no", you 
might get many answers to these and related questions from these 
references, including working the examples provided with the "glm" 
documentation -- and in reading the posting guide! 
"http://www.R-project.org/posting-guide.html". 

      hope this helps.  spencer graves

Kevin Cummins wrote:
>
> I have a system with a binary response variable that was hypothesized 
> to follow a simple logistic function. The relationship between the 
> continuous independent variable and the logit is clearly not 
> monotonic. I have two questions. 1) Can anyone recommend a reference 
> that describes my modeling options in this case, and 2) what 
> facilities does R have to deal with this situation?
>
> Thanks,
> Kevin Cummins
> School of Medicine
> University of California San Diego
>
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://www.stat.math.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide! 
> http://www.R-project.org/posting-guide.html