William M. Grove
2005-Apr-04 22:46 UTC
[R] R package that has (much) the same capabilities as SAS v9 PROC GENMOD
I need capabilities, for my data analysis, like the Pinheiro & Bates S-Plus/R package nlme() but with binomial family and logit link. I need multiple crossed, possibly interacting fixed effects (age cohort of twin when entered study, sex of twin, sampling method used to acquire twin pair, and twin zygosity), a couple of random effects other than the cluster variable, and the ability to have a variable of the sort that P&B call "outer" to the clustering variable---zygosity. Dependent variables are all parental (mom, dad separately of course) psychiatric diagnoses. In my data, twin pair ID is the clustering variable; correlations are expected to be exchangeable but substantially different between members of monozygotic twin pairs and members of dizygotic twin pairs. Hence, in my analyses, the variable that's "outer" to twin pair is monozygotic vs. dizygotic which of course applies to the whole pair. nlme() does all that but requires quasi-continuous responses, according to the preface/intro of P&B's mixed models book and what I infer from online help (i.e., no family= or link= argument). The repeated() library by Lindsey seems to handle just one nested random effect, or so I believe I read while scanning backlogs of the R-Help list. glmmPQL() is in the ballpark of what I need, but once again seems to lack the "outer" variable specification that nlme() has, and which PROC GENMOD also has---and which I need. I read someplace of yags() that apparently uses GEE to estimate parameters of nonlinear models including GLIMs/mixed models, just the way PROC GENMOD (and many another program) does. But on trying to install it (either v4.0-1.zip or v4.0-2.tar.gz from Carey's site, or Ripley's Windows port) from a local, downloaded zip file (or tar.gz file converted to zip file), I always get an error saying: > Error in file(file, "r") : unable to open connection > In addition: Warning message: > cannot open file `YAGS/DESCRIPTION' with no obvious solution. So I can't really try it out to see if it does what I want. You may ask: Why not just use GENMOD and skip the R hassles? Because I want to embed the GLIM/mixed model analysis in a stratified resampling bootstrapping loop. Very easy to implement in R, moderately painful to do in SAS. Can anybody give me a lead, or some guidance, about getting this job done in R? Thanks in advance for your help. Regards, Will Grove | Iohannes Paulus PP. II, xxx Psychology Dept. | U. of Minnesota | -----------------+ X-headers have PGP key info.; Call me at 612.625.1599 to verify key fingerprint before accepting signed mail as authentic! <br> <x-sigsep><p></x-sigsep> Will Grove | Iohannes Paulus PP. II, xxx <br> Psychology Dept. |<br> U. of Minnesota |<br> -----------------+<br> <br> X-headers have PGP key info.; Call me at 612.625.1599 to verify key fingerprint<br> before accepting signed mail as authentic!<br> <br> </body> </html>
Wensui Liu
2005-Apr-04 22:53 UTC
[R] R package that has (much) the same capabilities as SAS v9 PROC GENMOD
check glm() On Apr 4, 2005 6:46 PM, William M. Grove <grove001 at umn.edu> wrote:> I need capabilities, for my data analysis, like the Pinheiro & Bates > S-Plus/R package nlme() but with binomial family and logit link. > > I need multiple crossed, possibly interacting fixed effects (age cohort of > twin when entered study, sex of twin, sampling method used to acquire twin > pair, and twin zygosity), a couple of random effects other than the cluster > variable, and the ability to have a variable of the sort that P&B call > "outer" to the clustering variable---zygosity. Dependent variables are all > parental (mom, dad separately of course) psychiatric diagnoses. > > In my data, twin pair ID is the clustering variable; correlations are > expected to be exchangeable but substantially different between members of > monozygotic twin pairs and members of dizygotic twin pairs. Hence, in my > analyses, the variable that's "outer" to twin pair is monozygotic vs. > dizygotic which of course applies to the whole pair. > > nlme() does all that but requires quasi-continuous responses, according to > the preface/intro of P&B's mixed models book and what I infer from online > help (i.e., no family= or link= argument). > > The repeated() library by Lindsey seems to handle just one nested random > effect, or so I believe I read while scanning backlogs of the R-Help list. > > glmmPQL() is in the ballpark of what I need, but once again seems to lack > the "outer" variable specification that nlme() has, and which PROC GENMOD > also has---and which I need. > > I read someplace of yags() that apparently uses GEE to estimate parameters > of nonlinear models including GLIMs/mixed models, just the way PROC GENMOD > (and many another program) does. But on trying to install it (either > v4.0-1.zip or v4.0-2.tar.gz from Carey's site, or Ripley's Windows port) > from a local, downloaded zip file (or tar.gz file converted to zip file), I > always get an error saying: > > Error in file(file, "r") : unable to open connection > > In addition: Warning message: > > cannot open file `YAGS/DESCRIPTION' > with no obvious solution. > > So I can't really try it out to see if it does what I want. > > You may ask: Why not just use GENMOD and skip the R hassles? Because I > want to embed the GLIM/mixed model analysis in a stratified resampling > bootstrapping loop. Very easy to implement in R, moderately painful to do > in SAS. > > Can anybody give me a lead, or some guidance, about getting this job done > in R? Thanks in advance for your help. > > Regards, > > Will Grove | Iohannes Paulus PP. II, xxx > Psychology Dept. | > U. of Minnesota | > -----------------+ > > X-headers have PGP key info.; Call me at 612.625.1599 to verify key fingerprint > before accepting signed mail as authentic! > > <br> > <x-sigsep><p></x-sigsep> > Will Grove | Iohannes Paulus PP. II, > xxx <br> > Psychology Dept. |<br> > U. of Minnesota |<br> > -----------------+<br> > <br> > X-headers have PGP key info.; Call me at 612.625.1599 to verify key > fingerprint<br> > before accepting signed mail as authentic!<br> > <br> > </body> > </html> > > ______________________________________________ > R-help at 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 >-- WenSui Liu, MS MA Senior Decision Support Analyst Division of Health Policy and Clinical Effectiveness Cincinnati Children Hospital Medical Center