twostage: Two stage lmer boost

In patr1ckm/mvtboost: Tree Boosting for Multivariate Outcomes

Description

Fits a random intercept model with the grouping variable id to the residuals of gbm fit to the remaining predictors.

Usage

twostage(y, X, id, ..., cv.folds = 3, mc.cores = 1, subset = NULL)

Arguments

y	vector of observations of length n
X	matrix or data frame of predictors
id	name or index of grouping variable in x (can have missing values)
...	arguments passed to gbm_grid. Arguments can be passed as vectors as in
cv.folds	number of cross validation folds
mc.cores	number of cores to use in mclapply
subset	subset of observations tuning by cross validation will be carried out over expand.grid(...)

Details

The gbm model is cross-validated over a grid of all meta-parameters, as in gbm_grid. The model is fit to all predictors, except the grouping variable. Subsequently, a simple random intercept model (from lmer) is fit to the residuals, using id as a grouping variable, with the formula r ~ 1 + (1|id). A predict, and influence methods are provided for computing predictions and influence of predictors.

See Also

gbm_grid

patr1ckm/mvtboost documentation built on May 24, 2019, 8:21 p.m.