R/avgDensityBootstrap.R
avgDensityBootstrap.Rd
Compute bootstrap confidence intervals on the average squared density parameter
Compute bootstrap confidence intervals on the average squared density parameter
Output both the point estimate and the bootstrap confidence interval
TMLEbootstrap::generalBootstrap
-> avgDensityBootstrap
Inherited methods
new()
bootstrap average density parameter;
avgDensityBootstrap$new( x, epsilon_step = NULL, bin_width = 0.3, lambda_grid = NULL, M = NULL, targeting = TRUE )
x
vector of random samples.
epsilon_step
step size of the TMLE (one-step)
bin_width
binning width for the density super learner
lambda_grid
pre-specified grid of L1 penalization for the HAL
M
alternative to "lambda_grid". Specify the max variation norm of the HAL
targeting
bool. whether to perform targeting on the density estimator. If FALSE, the density super learner estimate will be returned
NULL
bootstrap_once()
run one bootstrap sample
avgDensityBootstrap$bootstrap_once( self, data, epsilon_step = self$epsilon_step, population_x = self$x, population_tmle = self$pointTMLE, inflate_lambda = 1 )
self
self
data
input bootstrap data
epsilon_step
step size of the TMLE (one-step)
population_x
whole training data
population_tmle
TMLE on the whole training data
inflate_lambda
experimental. scalar to scale the L1 penalty during the bootstrap
NULL
run_bootstrap()
avgDensityBootstrap$run_bootstrap( n_bootstrap = 200, alpha = 0.05, kind = NULL, inflate_lambda = 1, to_parallel = FALSE )
clone()
The objects of this class are cloneable with this method.
avgDensityBootstrap$clone(deep = FALSE)
deep
Whether to make a deep clone.