tvcm-control {vcrpart}

Control parameters for tvcm.

Description

Various parameters that control aspects for tvcm.

Usage

tvcm_control(minsize = 30, mindev = ifelse(sctest, 0.0, 2.0),
             sctest = FALSE, alpha = 0.05, bonferroni = TRUE,
             trim = 0.1, estfun.args = list(), nimpute = 5, 
             maxnomsplit = 5, maxordsplit = 9, maxnumsplit = 9,
             maxstep = 1e3, maxwidth = Inf, maxdepth = Inf,
             lossfun = neglogLik2, ooblossfun = NULL, fast = TRUE,
             cp = 0.0, dfpar = 0.0, dfsplit = 1.0,
             cv = !sctest, folds = folds_control("kfold", 5),
             prune = cv, papply = mclapply, papply.args = list(),
             center = fast, seed = NULL, verbose = FALSE, ...)

Arguments

alpha, bonferroni, trim, estfun.args, nimpute

See tvcolmm_control

mindev, cv, folds, prune, center

See tvcglm_control

minsize

numeric (vector). The minimum sum of weights in terminal nodes.

sctest

logical scalar. Defines whether coefficient constancy tests should be used for the variable and node selection in each iteration.

maxnomsplit

integer. For nominal partitioning variables with more the maxnomsplit the categories are ordered an treated as ordinal.

maxordsplit

integer. The maximum number of splits of ordered partitioning variables to be evaluated.

maxnumsplit

integer. The maximum number of splits of numeric partitioning variables to be evaluated.

maxstep

integer. The maximum number of iterations i.e. number of splits to be processed.

maxwidth

integer (vector). The maximum width of the partition(s).

maxdepth

integer (vector). The maximum depth of the partition(s).

lossfun

a function to extract the training error, typically minus two times the negative log likelihood of the fitted model (see neglogLik2). Is currently ignored if a glm model is fitted and fast = TRUE.

ooblossfun

a loss function that defines how to compute the validation error during cross-validation. The function will be assigned to the fun argument of oobloss.

fast

logical scalar. Whether the approximative model should be used to search for the next split. The approximative search model uses only the observations of the node to split and incorporates the fitted values of the current model as offsets. Therewith the estimation is reduces to the coefficients of the added split. If FALSE, the accurate search model is used.

cp

numeric scalar. The penalty to be multiplied with the complexity of the model during partitioning. The complexity of the model is defined as the number of coefficients times dfpar plus the number of splits times dfsplit. By default, cp = 0 (no penalization during partitioning) and dfpar = 0 and dfsplit = 1 (the complexity is measured as the total number of splits). cp also presents the minimum evaluated value at cross-validation.

dfpar

numeric scalar. The degree of freedom per model coefficient. Is used to compute the complexity of the model, see cp.

dfsplit

a numeric scalar. The degree of freedom per split. Is used to compute the complexity of the model, see cp.

papply

(parallel) apply function, defaults to mclapply. The function will parallelize the partition stage and the evaluation of the cross-validation folds as well as the final pruning stage.

papply.args

a list of arguments to be passed to papply.

seed

an integer specifying which seed should be set at the beginning.

verbose

logical. Should information about the fitting process be printed to the screen?

...

further, undocumented arguments to be passed.

Value

A list of class tvcm_control containing the control parameters for tvcm.

Author(s)

Reto Burgin

See Also

tvcolmm_control, tvcglm_control, tvcm, fvcm

Examples

tvcm_control(minsize = 100)