Hello everyone, I have tried using random forest as classifier for two classes. My data consists of 58 samples and each one of them belongs one of the two classes. When I tried runnng the random forest for the 58 samples, it gave me the followng error: Error in randomForest.default(m,y,...): Can not handle categorical predctors wth more than 32 categories. Although I have only 2 classes only. When i tried running it on 32 or less samples, it worked but when increased the samples, gave me this error. I thought maybe there is a limitations to the input data but the iris example has 150 samples and it works just fine. Any help? regards,salwa [[alternative HTML version deleted]]

On 03/06/2012 08:17 AM, Salwa Eid wrote: > > > > > Hello everyone, I have tried using random forest as classifier for two classes. My data consists of 58 samples and each one of them belongs one of the two classes. When I tried runnng the random forest for the 58 samples, it gave me the followng error: Error in randomForest.default(m,y,...): Can not handle categorical predctors wth more than 32 categories. Although I have only 2 classes only. When i tried running it on 32 or less samples, it worked but when increased the samples, gave me this error. I thought maybe there is a limitations to the input data but the iris example has 150 samples and it works just fine. Any help? regards,salwa > [[alternative HTML version deleted]] likely your data is not formatted correctly, perhaps confusing factor and level. But without a reproducible example it is hard to help. Martin > > _______________________________________________ > Bioconductor mailing list > Bioconductor at r-project.org > https://stat.ethz.ch/mailman/listinfo/bioconductor > Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor -- Computational Biology Fred Hutchinson Cancer Research Center 1100 Fairview Ave. N. PO Box 19024 Seattle, WA 98109 Location: M1-B861 Telephone: 206 667-2793

Hi, I was wondering if anybody could helping me in figuring out how to optimize explanatory variable selection using limma. Following is what I have been trying to do: library("limma") x.norm <- read.table("rma_data.txt", header=T, row.names=1, sep="\t") target <- read.table("target_wb01.txt", header=T, row.names=1, sep="\t") plasma <- as.factor(target$Plasma) stimulation <- as.factor(target$Stimulation) donor <- as.factor(target$Donor) design<- model.matrix(~ 0 + stimulation + plasma + donor) fit1<- lmFit(x.norm, design) fit2 <- contrasts.fit(fit1, cont.WT) fit3 <- eBayes(fit2) Is there any way to generate an anova table which gives me an idea as to which of the covariates; 'stimulation', 'plasma' or 'donor', in this case are informative and which aren't for all the probesets taken together. Any help would be greatly appreciated. Thanks. [[alternative HTML version deleted]]

Hi, On Tue, Mar 6, 2012 at 11:17 AM, Salwa Eid <salwaeid at="" hotmail.com=""> wrote: > > Hello everyone, ? ? I have tried using random forest as classifier for two classes. ?My data consists of 58 samples and each one of them belongs one of the two classes. ?When I tried runnng the random forest for the 58 samples, it gave me the followng error: Error in randomForest.default(m,y,...): ?Can not handle categorical predctors wth more than 32 categories. Although I have only 2 classes only. ?When i tried running it on 32 or less samples, it worked but when ?increased the samples, gave me this error. ? I thought maybe there is a limitations to the input data but the iris example has 150 samples and it works just fine. ? Any help? regards,salwa The error is telling you that you have some categorical variable/predictor (ie. one of the columns in your input data.frame that is *not* your label) that has more than 32 levels -- it's not talking about the number of classes (levels) your labels have. -- Steve Lianoglou Graduate Student: Computational Systems Biology ?| Memorial Sloan-Kettering Cancer Center ?| Weill Medical College of Cornell University Contact Info: http://cbio.mskcc.org/~lianos/contact

Hi Salwa, Please remember to "reply all" to keep this on the list. Try apply(mydata,2,class) to see the classes of the columns of the data matrix (mydata). I expect that somewhere you have another identifier column. --Naomi At 01:25 PM 3/6/2012, you wrote: > Yes I do have sample ids. I tried removing them and tried the > random forest but it still gave me the same error. If i use 32 or > less samples, it works fine but otherwise it gives me the error. > > Date: Tue, 6 Mar 2012 13:09:40 -0500 > > To: mtmorgan at fhcrc.org; salwaeid at hotmail.com > > From: naomi at stat.psu.edu > > Subject: Re: [BioC] error in using random forest > > CC: bioconductor at r-project.org > > > > Do you have sample ids? These could be the offending variables. > > > > --Naomi > > > > At 11:28 AM 3/6/2012, Martin Morgan wrote: > > >On 03/06/2012 08:17 AM, Salwa Eid wrote: > > >> > > >> > > >> > > >> > > >>Hello everyone, I have tried using random forest as classifier > > >>for two classes. My data consists of 58 samples and each one of > > >>them belongs one of the two classes. When I tried runnng the > > >>random forest for the 58 samples, it gave me the followng error: > > >>Error in randomForest.default(m,y,...): Can not handle categorical > > >>predctors wth more than 32 categories. Although I have only 2 > > >>classes only. When i tried running it on 32 or less samples, it > > >>worked but when increased the samples, gave me this error. I > > >>thought maybe there is a limitations to the input data but the iris > > >>example has 150 samples and it works just fine. Any help? > > >>regards,salwa > > >> [[alternative HTML version deleted]] > > > > > >likely your data is not formatted correctly, perhaps confusing > > >factor and level. But without a reproducible example it is hard > to help. Martin > > > > > >> > > >>_______________________________________________ > > >>Bioconductor mailing list > > >>Bioconductor at r-project.org > > >>https://stat.ethz.ch/mailman/listinfo/bioconductor > > >>Search the archives: > > >>http://news.gmane.org/gmane.science.biology.informatics.conductor > > > > > > > > >-- > > >Computational Biology > > >Fred Hutchinson Cancer Research Center > > >1100 Fairview Ave. N. PO Box 19024 Seattle, WA 98109 > > > > > >Location: M1-B861 > > >Telephone: 206 667-2793 > > > > > >_______________________________________________ > > >Bioconductor mailing list > > >Bioconductor at r-project.org > > >https://stat.ethz.ch/mailman/listinfo/bioconductor > > >Search the archives: > > >http://news.gmane.org/gmane.science.biology.informatics.conductor > > > > > >