wbonat
commented
5 years ago Hi, We have a bug here! Sorry! I will fix it asap. For now, I would suggest to perform the selection using univariate models instead of the multivariate one. The final result will be the same, because the selection is done for each response that's why you have to choose the argument response = 1.
x1 <- runif(100, -1, 1) x2 <- gl(2,50) beta1 = c(5, 0, 3) beta2 = c(15, 2, 1) X <- model.matrix(~ x1 + x2) y1 <- rnorm(100, mean = X%%beta1 , sd = 1) y2 <- rnorm(100, mean = X%%beta2 , sd = 1) data <- data.frame(y1, y2, x1, x2)
fit0 <- mcglm(c(y1 ~ 1), list(mc_id(data)), data = data) mc_sic(fit0, scope = c("x1", "x2"), data = data, response = 1)
hi and many many thanks for the package! I get this error when I run mc_sic on a multiresponse model: Error in inv_C %*% res : Cholmod error 'X and/or Y have wrong dimensions' at file ../MatrixOps/cholmod_sdmult.c, line 90
Reproducible code, slight modification of the example code in mc_sic man page. (I am running mcglm version 0.4.0, R version 3.5.2 on Ubuntu 16.04).
x1 <- runif(100, -1, 1) x2 <- gl(2,50) beta1 = c(5, 0, 3) beta2 = c(15, 2, 1) X <- model.matrix(~ x1 + x2) y1 <- rnorm(100, mean = X%%beta1 , sd = 1) y2 <- rnorm(100, mean = X%%beta2 , sd = 1) data <- data.frame(y1, y2, x1, x2) fit0 <- mcglm(c(y1 ~ 1, y2 ~ 1), list(mc_id(data), mc_id(data)), data = data) mc_sic(fit0, scope = c("x1", "x2"), data = data, response = 1)
error here
fiddling with mc_sic arguments does not help. Any solution? Many thanks!