On 13.03.2010 17:25, kloyt...@mappi.helsinki.fi wrote:
Lainaus "Uwe Ligges" <lig...@statistik.tu-dortmund.de>:On 05.03.2010 15:24, kloyt...@mappi.helsinki.fi wrote:Hi, I have a list p with different size dataframes and length of over 8000. I'm trying to calculate correlations between the rows of dataframes of this list and columns of another dataset (type data.frame also) so that first column is correlated with all the rows in the list dataframe. Some information from another dataset is also included to the final output (all.corrs). This worked a couple of weeks ago when I wrote it but now it seems not to, and gives an error message: Error in inherits(x, "data.frame") : subscript out of bounds In addition: There were 50 or more warnings (use warnings() to see the first 50) warnings() Warning messages: 1: In corrs[j] <- cbind(expressions[j, 1:5], SNP.expr.cor) : number of items to replace is not a multiple of replacement length which indicates that the problem is with getting correlation and other information into corrs. cbind(expressions[j,1:5], SNP.expr.cor) is type data.frame. Changing corrs into matrix, dataframe, list or any other type has not helped. I've updated R from 2.9.0 to the recent version in between. Would anyone have a solution for this problem? I very much appreciate all help. SNP.expr.cor<-NULL all.corrs<-NULL corrs<-NULL for(i in 1:length(p)){ dim.exp<-dim(p[[i]]) expressions<-p[[i]] expressions.m<-as.matrix(expressions[,6:48]) for(j in 1:dim.exp[1]){ SNP.expr.cor<-cor(genotypes[,i],expressions.m[j,],use="na.or.complete") corrs[j]<-cbind(expressions[j,1:5], SNP.expr.cor) } all.corrs[i]<-list(cbind(map[i,1:6], corrs)) } BR KatjaYour example is not reproducible, since we do not have the data. What I guess is that you need to make your objects lists in advance. I cannot try out, but maybe the following works better right away all.corrs <- vector(mode = "list", length = length(p)) for(i in seq(along = p)){ dim.exp <- dim(p[[i]]) corrs <- vector(mode = "list", length = dim.exp[1]) expressions <- p[[i]] expressions.m <- as.matrix(expressions[,6:48]) for(j in 1:dim.exp[1]){ SNP.expr.cor <- cor(genotypes[, i], expressions.m[j, ], use = "na.or.complete") corrs[[j]] <- cbind(expressions[j, 1:5], SNP.expr.cor) } all.corrs[[i]] <- list(cbind(map[i, 1:6]), corrs) } Best, Uwe LiggesHi, sorry about that, here's a new try: corrs<-NULL #I've tried everything from vector to matrix to data.frame to list a<-rnorm(30) b<-cbind(a,a,a,a,a) e<-as.matrix(t(b)) c<-c("one", "two", "three", "four", "five") d<-c("one", "two", "three", "four", "five") c<-rbind(c,c,c,c,c) k<-dim(dat)
Well, dat is still unknown to us.
for(j in 1:k[1]){
dat.cor<-cor(b[,i],e[j,],use="na.or.complete")
i is unknown to us now. Your code is still *not* reproducible (note that I forgot what happened a week ago). Please post your whole reproducible code! We do not remember all things for weeks....
corrs[j]<-cbind(c[j,1:5], dat.cor)
That cannnot work: corrs is a vector (given you said nothing else and given your indexing). You cannot assign cbind(c[j,1:5], dat.cor) which is a matrix of 6 columns at least, into a single element of a vector corrs[j].
If your dat.cor and you c are numeric values (rather than more complex objects that I believe they are intended to be from your former message), then you caould assign into rows of a matrix corrs by initialiszing it before and assigning rowwise by
corrs[j,] <- cbind(c[j,1:5], dat.cor)But since you do not know this basic assigment trick, I really doubt you know how complex your objects really are... It might make sense to learn a bit more of R and reading some R introducing material as well as asking a local expert.
Best, Uwe Ligges
} The trouble seems to be in the inner loop. I'd love to have it as data.frame or matrix, not as a list. I think there has been similar problems presented here before but I haven't been able to find a working solution yet. BR Katja Löytynoja
______________________________________________ R-help@r-project.org 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.
