Boosting algorithms for fitting generalized linear, additive and interaction models to potentially high-dimensional data. The current relase version can be found on CRAN (http://cran.r-project.org/package=mboost).
73
stars
27
forks
source link
Cross validation (CVRisk) using a Cox PH model #95
I've came to notice that in version 2.9.0, cross validation for Cox PH models when corrected = True and Leave-One-Out cross validation when corrected = False were deleted. Nonetheless, I believe cross-validation for Cox PH models with k-fold cross validation should not be affected by the update.
In the file crossvalidation.R, line 38-39 currently prints the following code:
if (fam_name == "Cox Partial Likelihood" && all(rowSums(folds == 0) == 1))
stop("Leave-one-out cross-validation cannot be used with ", sQuote("family = CoxPH()"))
However, rows here represent the samples while columns represent the different folds used in cross-validation. The current implementation would always stop regardless of the number of folds specified in cross-validation. Shouldn't the stop criterion be all(colSums(folds==0)==1) to only exclude Leave-One-Out cross-validation?
I've came to notice that in version 2.9.0, cross validation for Cox PH models when corrected = True and Leave-One-Out cross validation when corrected = False were deleted. Nonetheless, I believe cross-validation for Cox PH models with k-fold cross validation should not be affected by the update.
In the file crossvalidation.R, line 38-39 currently prints the following code:
if (fam_name == "Cox Partial Likelihood" && all(rowSums(folds == 0) == 1)) stop("Leave-one-out cross-validation cannot be used with ", sQuote("family = CoxPH()"))
However, rows here represent the samples while columns represent the different folds used in cross-validation. The current implementation would always stop regardless of the number of folds specified in cross-validation. Shouldn't the stop criterion be all(colSums(folds==0)==1) to only exclude Leave-One-Out cross-validation?