SFM.bootstrap: SFM.bootstrap performs estimation with B Individual Bootstrap...

In clemenshaerder/fepsfrontieR: Estimation of fixed-effect panel stochastic frontier models by MLE

Description

B Individual Bootstrap Samples are generated from the input and MLE is performed for each sample. Unlike i.i.d. bootstrapping, individual bootrapping samples the rows with replacement individually for each panel instead from all panels. In addition to the mean and the standard error of the estimates, a confidence interval is returned based on the quantiles of the distribution of estimates.

Usage

SFM.bootstrap(y, xv, z, mu, N, Time, method, R, K, B, myPar = NULL,
  lowerInt, alphaCI, cumTime, parallel)

Arguments

y	is a n*t x 1 vector (response)
xv	is a n*t x k matrix (explantatory variables)
z	is a n*t x r matrix (inefficency determinants)
mu	is an integer (mean of the truncated normal distribution of the inefficency)
N	is an integer (n - panels)
Time	is an integer (observations per panel)
method	a required string specifying the method ("within" or "firstdiff").
R	is an integer (# of z variables)
K	is an integer (# of xv variables)
B	is an integer (# of Bootstraps)
myPar	is a vecor which has to be entered in the following order: c(sigma_v, sigma_u, beta = c(), delta = c()). Required as starting point for the estimation.
lowerInt	is a vector of doubles (lower bound for the estimation)
alphaCI	is a vector of doubles (significance of the Confidence Intervals)
cumTime	ia a vector of the cumulated times of the Time vector. It serves as an index for computation.
parallel	is an optional boolean variable. If it is set to TRUE, bootstrapping is performed with parallelization, using all available cores - 1. Only available for OS Windows.

Value

A B x (K + R + 2) matrix is returned of the estimates, the mean, standard error and a confidence interval for each estimate as a data frame.

clemenshaerder/fepsfrontieR documentation built on May 22, 2019, 3:43 p.m.