Thanks all. I preferred not to write a modified pamr.confusion (even if I think a "return" of the table should be included by the maintainers of pamr package...or not?). Setting "extra" parameter only removes error rates, but the resulting reduced table then is writeable as a file. Anyway I would like to have the complete table with errors too, and so diverting the output via "sink" works best for me... !!! Thanks a lot !!! Giulio >From: Bj?rn Usadel <usadel at="" mpimp-golm.mpg.de=""> >To: Giulio Di Giovanni <perimessaggini at="" hotmail.com="">, >bioconductor at stat.math.ethz.ch >Subject: Re: [BioC] How to write a text file of a pamr.confusion table ? >Date: Tue, 29 Nov 2005 16:05:11 +0100 > >Hi Giulio, > >you might try sink for diverting the output. > >Otherwise,.... >Having a look at pamr.confusion (> pamr.confusion on the command line) >shows that only if the parameter extra is set to false you get a confusion >matrix returned. >So either you set it extra to False, or you redefine the function > >pamr.Konfusion<- function (fit, threshold, extra = TRUE) >{ > ii <- (1:length(fit$threshold))[fit$threshold >= threshold] > ii <- ii[1] > predicted <- fit$yhat[, ii] > if (!is.null(fit$y)) { > true <- fit$y[fit$sample.subset] > tt <- table(true, predicted) > } > else { > true <- fit$proby[fit$sample.subset, ] > ytemp <- apply(true, 1, which.is.max) > temp <- c(predicted, names(table(ytemp))) > nams <- names(table(temp)) > Yhat <- model.matrix(~factor(temp) - 1, data = list(y = temp)) > Yhat <- Yhat[1:length(predicted), ] > tt <- matrix(NA, nrow = length(fit$prior), ncol = >length(fit$prior)) > for (i in 1:length(fit$prior)) { > for (j in 1:length(fit$prior)) { > tt[i, j] <- sum(true[, i] * Yhat[, j]) > } > } > dimnames(tt) <- list(names(table(ytemp)), nams) > } > >if (extra) { > tt1 <- tt > diag(tt1) <- 0 > tt <- cbind(tt, apply(tt1, 1, sum)/apply(tt, 1, sum)) > dimnames(tt)[[2]][ncol(tt)] <- "Class Error rate" > print(tt) #<-****can go > cat(c("Overall error rate=", round(sum(tt1)/sum(tt), > 3)), fill = TRUE) > return(tt) # <-**** add this >} > if (!extra) { > return(tt) > } >} > > >Cheers, >Bj?rn > >>Hi, >> >>really I hope this is not a stupid question. I would like to write a text >>file with the output of pamr.confusion function. Obviously I tried first >>with the usual >> >> >> >>>write.table(file="test.txt",pamr.confusion(mysamplespamtrain, thr)) >>> >>> >> K N Class Error rate >>K 26 4 0.1333333 >>N 2 23 0.0800000 >>Overall error rate= 0.109 >> >>but all I write is an empty file. The same with "write" or assigning >>pamr.confusion to a variable (pc <- pamr.confusion(.....) ). >>I also tried changing the device, to write a jpeg or whatelse; I tried to >>play with Device related commands. No way. I followed James McDonald >>suggestion >>found in the mail archive. >> >> >> >>>pamr.confusion(mData.results, threshold=4.0) >>> >>> >>(table) >> >> >>>savePlot("Confusion", "wmf") >>> >>> >>Error in savePlot("Confusion", "wmf") : no such device (...well, there's >>no graph device...). >> >>Anyway, there's anyone who has some hints ? I think that maybe is a simple >>command, but I don't know how to do it. >> >>Thanks in advance, and sorry if is a too banal question... >> >>Giulio >> >>_______________________________________________ >>Bioconductor mailing list >>Bioconductor at stat.math.ethz.ch >>https://stat.ethz.ch/mailman/listinfo/bioconductor >> >> >