Re: [R] glmmPQL question!

From: Spencer Graves <spencer.graves_at_pdf.com>
Date: Thu 17 Aug 2006 - 09:57:01 EST

  I don't think there is an easy way to do that. If it were my problem, I think I'd start by trying to the model using 'lmer' associated with the 'lme4' package. I would then try to pass the fit to 'mcmcsamp' to get a random sample of parameter estimates following the posterior distribution. Then I'd convert the simulated parameters into a distribution of predictions for the conditions of interest. While doing this, I'd apply "quantile" of the distribution of predictions for each set of conditions to get the desired confidence limits.

Make sense?

If you'd like further help from this listserve, please submit another post. When you do that, however, I strongly urge you to include commented, minimal, self-contained, reproducible code illustrating something you tried that didn't quite work to help people understand what you want (as suggested in the posting guide "www.R-project.org/posting-guide.html"). Without such a minimal, self contained example, your problem is rarely as clear, and people's replies are less likely to be relevant. Since they know that, they are less likely to reply.

Hope this helps.
Spencer Graves

m-krutky@northwestern.edu wrote:
> Hello Folks-
>
> Is there a way to create confidence bands with 'glmmPQL' ???
>
> I am performing a stroke study for Northwestern University in Chicago, Illinois. I am trying to
> decide a way to best plot the model which we created with the glmmPQL function in R. I would like
> to plot my actual averaged data points within 95 % confidence intervals from the model. Plotting
> the model is easy, but determining confidence bands is not.
>
> Here is my model:
>
> ratiomodel<-glmmPQL(ratio~as.factor(joint)*time, random = ~ 1 | subject, family = Gamma(link =
> "identity"),alldata3)
>
> I am used to seeing confidence intervals from models that increase, “flair out” in the y direction,
> at the beginning and ending time points (x values) of the simulated data. If I use 'lm' and pass
> the command 'int = "c" ' 'to create this model I can easily find and plot this type of confidence
> band for 'ratio~time'. But I need to take into account 'as.factor(joint)', and in fact I can
> produce confidence bands with 'glm' by passing in 'se.fit = TRUE', but the problem is I need to
> make subject a random variable, and take into account my ratio with the Gamma distribution.
>
> Is there a way to create confidence bands with 'glmmPQL' ??? ' as.factor(joint)' has 3 levels, so I
> would like to produce this linear model with three levels and confidence bands for comparison of
> the levels of 'joint'.
>
> Any Help at all with my problem would be greatly appreciated!!
> LJ
>
>
> ------------------------------------------------------------------------
>
> ______________________________________________
> 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
> and provide commented, minimal, self-contained, reproducible code.
>



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 and provide commented, minimal, self-contained, reproducible code. Received on Thu Aug 17 10:10:44 2006

Archive maintained by Robert King, hosted by the discipline of statistics at the University of Newcastle, Australia.
Archive generated by hypermail 2.1.8, at Thu 17 Aug 2006 - 20:23:11 EST.

Mailing list information is available at https://stat.ethz.ch/mailman/listinfo/r-help. Please read the posting guide before posting to the list.