mermboost: Gradient Boosting for Additive Mixed Models

In mermboost: Gradient Boosting for Generalized Additive Mixed Models

Gradient Boosting for Additive Mixed Models

Description

Gradient boosting for optimizing negative log-likelihoods as loss functions, where component-wise arbitrary base-learners, e.g., smoothing procedures, are utilized as additive base-learners. In addition, every iteration estimates random component via a maximum likelihood approach using the current fit.

Usage

mermboost(formula, data = list(), na.action = na.omit, weights = NULL,
       offset = NULL, family = gaussian, control = boost_control(),
       oobweights = NULL, baselearner = c("bbs", "bols", "btree", "bss", "bns"),
       ...)

Arguments

formula	a symbolic description of the model to be fit in the lme4-format including random effects.
data	a data frame containing the variables in the model.
na.action	a function which indicates what should happen when the data contain NAs.
weights	(optional) a numeric vector of weights to be used in the fitting process.
offset	a numeric vector to be used as offset (optional).
family	!! This is in contrast to usual mboost - "only" a family object is possible - except for NBinomial().
control	a list of parameters controlling the algorithm. For more details see boost_control.
oobweights	an additional vector of out-of-bag weights, which is used for the out-of-bag risk (i.e., if boost_control(risk = "oobag")). This argument is also used internally by cvrisk.
baselearner	a character specifying the component-wise base learner to be used: bbs means P-splines with a B-spline basis (see Schmid and Hothorn 2008), bols linear models and btree boosts stumps. bss and bns are deprecated. Component-wise smoothing splines have been considered in Buehlmann and Yu (2003) and Schmid and Hothorn (2008) investigate P-splines with a B-spline basis. Kneib, Hothorn and Tutz (2009) also utilize P-splines with a B-spline basis, supplement them with their bivariate tensor product version to estimate interaction surfaces and spatial effects and also consider random effects base learners.
...	additional arguments passed to mboost_fit; currently none.

Details

A (generalized) additive mixed model is fitted using a boosting algorithm based on component-wise base-learners. Additionally, a mixed model gets estimated in every iteration and added to the current fit.

The base-learners can either be specified via the formula object or via the baselearner argument. The latter argument is the default base-learner which is used for all variables in the formula, without explicit base-learner specification (i.e., if the base-learners are explicitly specified in formula, the baselearner argument will be ignored for this variable).

Of note, "bss" and "bns" are deprecated and only in the list for backward compatibility.

Note that more base-learners (i.e., in addition to the ones provided via baselearner) can be specified in formula. See baselearners for details.

Value

The description of mboost holds while some methods are newly implemented like predict.mermboost, plot.mer_cv and mstop.mer_cv. Only the former one requires an further argument. Additionally, methods VarCorr.mermboost and ranef.mermboost are implemented specifically.

See Also

See glmermboost for the same approach using additive models.

See mer_cvrisk for a cluster-sensitive cross-validation.

Examples

data("Orthodont")

# are there cluster-constant covariates?
find_ccc(Orthodont, "Subject")


mod <- mermboost(distance ~ bbs(age, knots = 4) + bols(Sex) + (1 |Subject),
                 data = Orthodont, family = gaussian,
                 control = boost_control(mstop = 100))

# let mermboost do the cluster-sensitive cross-validation for you
norm_cv <- mer_cvrisk(mod, no_of_folds = 10)
opt_m <- mstop(norm_cv)

# fit model with optimal stopping iteration
mod_opt <- mermboost(distance ~ bbs(age, knots = 4) + bols(Sex) + (1 |Subject),
                 data = Orthodont, family = gaussian,
                 control = boost_control(mstop = opt_m))

# use the model as known from mboost
# in additional, there are some methods knwon from lme4
ranef(mod_opt)
VarCorr(mod_opt)

mermboost documentation built on April 4, 2025, 1:41 a.m.