BoostMLR (version 1.0.3)

vimp.BoostMLR: Variable Importance

Description

Calculate standardized variable importance (VIMP) for each covariate or a joint VIMP of multiple covariates.

Usage

# S3 method for BoostMLR
vimp(Object,
     xvar.names = NULL,
     joint = FALSE,
     setting_seed = FALSE,
     seed_value = 100L)

Arguments

Object

A boosting object of class (BoostMLR, predict).

xvar.names

Names of the x-variables for which VIMP is requested. If NULL, VIMP is calcuated for all the covariates.

joint

Whether to estimate VIMP for each covariate from xvar.names or a joint VIMP for multiple covariates?

setting_seed

Set setting_seed = TRUE if you intend to reproduce the result.

seed_value

Seed value.

Value

If joint = FALSE, a standardized VIMP for each covariate is obtained otherwisea joint VIMP for all the covariates is obtained. The result consists of a list of length equal to the number of multivariate response. Each element from the list represents a matrix with number of rows equal to the number of covariates (in case of joint VIMP, the matrix will have a single row) and the number of columns equal to the number of overlapping time intervals + 1 where the first column contains covariate main effects and all other columns contain covariate-time interaction effects.

Details

Standardized variable importance (VIMP) is calcuated for each covariate or a joint VIMP is calculated for all the covariates specified in xvar.names.

References

Pande A., Ishwaran H., Blackstone E.H. (2020). Boosting for multivariate longitudinal response.

Friedman J.H. Greedy function approximation: a gradient boosting machine, Ann. of Statist., 5:1189-1232, 2001.

Examples

# NOT RUN {
##-----------------------------------------------------------------
## Calculate individual and joint VIMP
##-----------------------------------------------------------------

# Simulate data involves 3 response and 4 covariates

dta <- simLong(n = 100, ntest = 100 ,N = 5, rho =.80, model = 1, q_x = 0, 
                                  q_y = 0,type = "corCompSym")
dtaL <- dta$dtaL
trn <- dta$trn
# Boosting call: Raw values of covariates, B-spline for time, 
# no shrinkage, no estimate of rho and phi

boost.grow <- BoostMLR(x = dtaL$features[trn,], tm = dtaL$time[trn], 
                      id = dtaL$id[trn], y = dtaL$y[trn,], M = 100, VarFlag = FALSE)

boost.pred <- predictBoostMLR(Object = boost.grow, x = dtaL$features[-trn,], 
                               tm = dtaL$time[-trn], id = dtaL$id[-trn], 
                               y = dtaL$y[-trn,], importance = FALSE)
# Individual VIMP                               
Ind_vimp <- vimp.BoostMLR(boost.pred)

# Joint VIMP
Joint_vimp <- vimp.BoostMLR(boost.pred,joint = TRUE)

# }