Exactly Bill! I'll implement your suggestion (solve_svd(cor.x2)%*%cor.y), to return betas. Best! Tiago
Em sex., 4 de jun. de 2021 às 15:12, Bill Dunlap <williamwdun...@gmail.com> escreveu: > solve_svd's second argument is 'tolerance', a small scalar value, not the > right hand side of X %*% beta == Y. It returns the [approximate] inverse > of X, not beta. > > solve_svd(cor.x2,cor.y) should probably be solve_svd(cor.x2)%*%cor.y (or > just solve(cor.x2,cor.y)). > > -Bill > > On Fri, Jun 4, 2021 at 8:41 AM Bill Dunlap <williamwdun...@gmail.com> > wrote: > >> The offending line in path_coeff seems to be >> betas[i, 2:nvar] <- t(solve_svd(cor.x2, cor.y)) >> where i is a single integer, nvar is 15, and the right hand side is a 14 >> by 14 matrix. What is this line trying to do? Did it ever give the >> correct result? >> >> -Bill >> >> >> On Fri, Jun 4, 2021 at 7:39 AM Bill Dunlap <williamwdun...@gmail.com> >> wrote: >> >>> That log file includes the line >>> using R Under development (unstable) (2021-05-30 r80413) >>> and later says >>> >>> The error most likely occurred in: >>> >>> > ### Name: path_coeff >>> > ### Title: Path coefficients with minimal multicollinearity >>> > ### Aliases: path_coeff path_coeff_mat >>> > >>> > ### ** Examples >>> > >>> > ## No test: >>> > library(metan) >>> > >>> > # Using KW as the response variable and all other ones as predictors >>> > pcoeff <- path_coeff(data_ge2, resp = KW) >>> Error in matrix(value, n, p) : >>> data length differs from size of matrix: [196 != 1 x 14] >>> Calls: path_coeff -> [<- -> [<-.data.frame -> matrix >>> Execution halted >>> >>> In the current R-devel the matrix command matrix(value, nrow, ncol) >>> complains if length(value)>nrow*ncol, so you need to truncate 'value'. >>> (Previously matrix() would silently truncate value.) >>> >>> >>> -Bill >>> >>> >>> On Thu, Jun 3, 2021 at 5:57 PM Tiago Olivoto <tiagooliv...@gmail.com> >>> wrote: >>> >>>> Dear all, >>>> I have received an email from the CRAN team to fix a problem with my R >>>> package metan <https://CRAN.R-project.org/package=metan >>>> <https://cran.r-project.org/package=metan>> to safely retain it on >>>> CRAN. >>>> >>>> The error is given at < >>>> https://www.stats.ox.ac.uk/pub/bdr/donttest/metan.out> >>>> and seems to be related to data different from size of matrix. >>>> >>>> The question is that and I cannot reproduce the error locally and thus >>>> not >>>> able to fix anything. By googling the error I've seen similar issues >>>> with a >>>> lot of other packages. Could be this be a temporary issue? >>>> >>>> Thanks in advance! >>>> Tiago >>>> >>>> [[alternative HTML version deleted]] >>>> >>>> ______________________________________________ >>>> R-package-devel@r-project.org mailing list >>>> https://stat.ethz.ch/mailman/listinfo/r-package-devel >>>> >>> [[alternative HTML version deleted]] ______________________________________________ R-package-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-package-devel
