When fitting an FDboostLSS model with method = "noncyclic" the resulting S3-object has the classes nc_mboostLSS, FDboostLSS and mboostLSS in this order. Thus, the defaults specified in FDboost:::cvrisk.FDboostLSS won't apply. (See code below.)
Two quick ideas would be to A) change the class order, if that's unproblematic, or B), stress that the folds have to be specified manually in the documentation and deprecate FDboost:::cvrisk.FDboostLSS.
library(FDboost)
library(gamboostLSS)
########### simulate Gaussian function-on-scalar data
n <- 50 ## number of observations
G <- 12 ## number of observations per functional covariate
set.seed(123) ## ensure reproducibility
n_innerknots <- 4
B <- mboost:::bsplines(1:G, knots = 4, boundary.knots = c(1,G), degree = 2)
theta <- rnorm(ncol(B)) ## sample coefficients for x = 1
x <- runif(n) ## sample covariates
y <- B %*% matrix( rnorm(n*length(theta), mean = rep(theta, n) * rep(x, each = length(theta)), sd = rep(x, each = length(theta))), ncol = n) ## sample response
dat_list <- list(y = t(y), x = x, t = 1:G)
## model fit assuming Gaussian location scale model
model <- FDboostLSS(formula = y ~ bols(x, df = 2),
timeformula = ~ bbs(t, df = 2),
data = dat_list, method = "noncyclic")
class(model)
## -> cvrisk.nc_mboostLSS is directly applied
debug(cvrisk)
cvrisk(model)
RetroSearch is an open source project built by @garambo | Open a GitHub Issue
Search and Browse the WWW like it's 1997 | Search results from DuckDuckGo
HTML:
3.2
| Encoding:
UTF-8
| Version:
0.7.4