Dear all I have a big matrix of standardized values (dimensions 285x5829) and R fails to calculate the principal components using prcomp() with the following error message: pc <- prcomp(my.matrix) Error in La.svd(x, nu, nv, method) : error code 1 from Lapack routine dgesdd Is the matrix too big? I'm using R-1.8.1 under Unix (Solaris8) and Linux(Suse 8.2). I tried to perform a principal component analysis using a huge matrix with the dimensions 1000x10000 and it worked without an error message. test <- matrix(rnorm(10000000), ncol=1000, nrow=10000) pc <- prcomp(test) Any hints are welcome. Many thanks in advance and best regards, Carlo ****************************************************** Carlo Casty Climatology and Meteorology Phone: +41 (0)31 6318545 Institute of Geography Fax: +41 (0)31 6318511 University of Bern e-mail:casty at giub.unibe.ch Hallerstr. 12 www.giub.unibe.ch/~casty 3012 Bern/SWITZERLAND
Prof Brian Ripley
2004-Mar-04 11:03 UTC
[R] prcomp: error code 1 from Lapack routine dgesdd
On Thu, 4 Mar 2004, Carlo Casty wrote:> Dear all > > I have a big matrix of standardized values (dimensions 285x5829) and R > fails to calculate > the principal components using prcomp() with the following error message: > > pc <- prcomp(my.matrix) > Error in La.svd(x, nu, nv, method) : error code 1 from Lapack routine > dgesdd > > Is the matrix too big? I'm using R-1.8.1 under Unix (Solaris8) andNo. It indicates a failure of the algorithm used, hence something about the matrix in your example.> Linux(Suse 8.2). I tried to > perform a principal component analysis using a huge matrix with the > dimensions 1000x10000 and it worked without an error message. > > test <- matrix(rnorm(10000000), ncol=1000, nrow=10000) > pc <- prcomp(test)-- Brian D. Ripley, ripley at stats.ox.ac.uk Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/ University of Oxford, Tel: +44 1865 272861 (self) 1 South Parks Road, +44 1865 272866 (PA) Oxford OX1 3TG, UK Fax: +44 1865 272595