grridge model fitting and predicting for survival data fails

[HerrMo]

Hi,

I'm trying to get grridge working for survival data and encounter some problems when fitting the model and predicting new observation. Below you find reproducible examples. First of all, in Examples 1 and 2 there is exactly the same code to fit a GRridge model. The only thing that changes is the seed. Unfortunately, only Example 1 works, Example 2 fails and throws an error. Secondly, in Example 3, a model is fit with the same code as in Example 1, but then predicting new observations fails.

library(GRridge)
library(prodlim)

# Example 1: fitting works --------------------------------------------------------

set.seed(4) 
n_col <- 200

# random design matrix 
X <- cbind(matrix(nrow=40, ncol=n_col, data=rnorm(40*n_col)),
           matrix(nrow=40, ncol=30, data=rnorm(40*30, mean=1, sd=2)),
           matrix(nrow=40, ncol=100, data=rnorm(40*100, mean=2, sd=3)))

# grouping
blocks <- rep(1:3, times=c(n_col, 30, 100))
blocks <- lapply(1:3, function(x) which(blocks==x))

# survival outcome
ysurv <- prodlim::SimSurv(40)[, c(3,7)]

# split in test and train set
n <- sample.int(nrow(X), 2/3*nrow(X))

train_x <- t(X[n, ])
test_x <- t(X[-n, ])

# fit 
tt <- GRridge::grridge(train_x,
                       Surv(ysurv[n, 1], ysurv[n, 2]),
                       partitions = blocks,
                       standardizeX = TRUE,
                       innfold = 5,
                       selectionEN = TRUE)

# Example 2: fitting does not work ----------------------------------------------------------

# same code as above except for seed
rm(list = ls())

set.seed(16) 

n_col <- 200

# random design matrix 
X <- cbind(matrix(nrow=40, ncol=n_col, data=rnorm(40*n_col)),
           matrix(nrow=40, ncol=30, data=rnorm(40*30, mean=1, sd=2)),
           matrix(nrow=40, ncol=100, data=rnorm(40*100, mean=2, sd=3)))

# grouping
blocks <- rep(1:3, times=c(n_col, 30, 100))
blocks <- lapply(1:3, function(x) which(blocks==x))

# survival outcome
ysurv <- prodlim::SimSurv(40)[, c(3,7)]

# split in test and train set
n <- sample.int(nrow(X), 2/3*nrow(X))

train_x <- t(X[n, ])
test_x <- t(X[-n, ])

# fit 
tt <- GRridge::grridge(train_x,
                       Surv(ysurv[n, 1], ysurv[n, 2]),
                       partitions = blocks,
                       standardizeX = TRUE,
                       innfold = 5,
                       selectionEN = TRUE)

Error in qr.coef(qr(a, LAPACK = TRUE), b) : 'qr' and 'y' must have the same number of rows Error: Matrix inversion failed. Please increase lambda1 and/or lambda2

# Example 3: prediction fails ------------------------------------------------------------

# Code is the same as in Example 1 except 'standardizeX'-argument and prediction 
rm(list = ls())

set.seed(4) 

n_col <- 200

# random design matrix 
X <- cbind(matrix(nrow=40, ncol=n_col, data=rnorm(40*n_col)),
           matrix(nrow=40, ncol=30, data=rnorm(40*30, mean=1, sd=2)),
           matrix(nrow=40, ncol=100, data=rnorm(40*100, mean=2, sd=3)))

# grouping
blocks <- rep(1:3, times=c(n_col, 30, 100))
blocks <- lapply(1:3, function(x) which(blocks==x))

# survival outcome
ysurv <- prodlim::SimSurv(40)[, c(3,7)]

# split in test and train set
n <- sample.int(nrow(X), 2/3*nrow(X))

train_x <- t(X[n, ])
test_x <- t(X[-n, ])

# fit 
tt <- GRridge::grridge(train_x,
                       Surv(ysurv[n, 1], ysurv[n, 2]),
                       partitions = blocks,
                       standardizeX = FALSE, 
                       innfold = 5,
                       selectionEN = TRUE)

prd_tt <- predict.grridge(tt, test_x)

Error in dimnames(x) <- dn : length of 'dimnames' [2] not equal to array extent

The data in the examples is generated randomly due to reproducibility, but I encounter the same problems when I use real data.

Any help would be highly appreciated. Thanks

[markvdwiel]

Thanks for your interest. The first issue was due to an Infinite penalty, the second a genuine error. Both issues should now have been resolved. Please re-install GRridge using library(devtools) install_github("markvdwiel/GRridge")

[HerrMo]

Thank you very much, it is working now!