sfmfep: Fixed-Effect Stochastic Frontier Model Estimation for Panel...

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

Description

Estimates a Stochastic Frontier Model for Fixed-Effects. sfmfep is used to fit fixed-effect stochastic frontier models for panel data using a specified model transformation. Bootstrapping can be performed to calculate the standard errors instead of a numerical deriviation via the hessian matrix.

Usage

sfmfep(formula, data, panel = NULL, N = NULL, Time = NULL,
  method = "firstdiff", mu = 0, alphaCI = 0.05, estimate = T,
  bootstrap = F, B = NULL, parallel = F, myPar = c(sigma_u = NULL,
  sigma_v = NULL, beta = c(NULL), delta = c(NULL)))

Arguments

formula	an object of class "formula" in the form of y ~ x1 + ... + x_k + (z1 + ... + z_r)
data	an optional data frame, list or environment (or object coercible by as.data.frame to a data frame) containing the variables in the model.
panel	an optional vector specifying the panels to be used in the fitting process.
N	an optional integer specifying the total amount of panels in the data set. If N is entered, Time is also a required input.
Time	an optional integer specifying the amount of observations per panel. If Time is entered, N is also a required input.
method	a required string specifying the method ("within" or "firstdiff").
mu	is the mean of a truncated normal distribution of the stochastic inefficiency.
alphaCI	is an optional vector specifying the significance values of the confidence intervals for the MLE estimates.
estimate	TRUE or FALSE specifies if "myPar" is used as starting point of the estimation, or if "myPar" is used to fit a given stochastic frontier model.
bootstrap	is an optional boolean variable. If it is set to TRUE, bootstrapping is performed. "B" must be specified.
B	is an integer which is a required input for Bootstrap = T. It defines the amount of bootstrap samples.
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.
myPar	is a vecor which has to be entered in the following order: c(sigma_v, sigma_u, beta = c( ), delta = c( ))

Value

sfmfep returns an object of class S3 "sfmfep". The function summary( ) can be used to obtain or print a summary of the results.

Examples

# Fit of a simple model with balanced panels.
# Definition with *N & T*.
# Data has 10 observations for each of the 20 panels.

fit1 <- sfmfep(formula = y ~ x + (z),
    method = "within", N = 40, Time = 5, data = sfm.data)
summary(fit1)

# ---------------
# Fit of a simple model with balanced panels.
# Definition with *panels*.

fit2 <- sfmfep(formula = y ~ x + (z),
     method = "within", panel = sfm.data$producer, data = sfm.data)
summary(fit2)

# ---------------
# Fit of a simple model with *Bootstrapping* using
# *method = firstdiff*
# with different *sigmas* for *Confidence Intervals*

fit3 <- sfmfep(formula = y ~ x + (z),
    bootstrap = TRUE, B = 20, alphaCI = c(0.1, 0.05),
    method = "firstdiff", panel = sfm.data$producer, data = sfm.data)
summary(fit3)

# ---------------
# Fitting a model *without estimating*, providing model parameters.

fit4 <- sfmfep(formula = y ~ x + (z),
    myPar = c(sigma_u = 0.2, sigma_v = 0.1, beta = 0.5, delta = 0.5),
    estimate = FALSE,
    method = "firstdiff", panel = sfm.data$producer, data = sfm.data)
summary(fit4)

# ---------------
# Perform Bootstrapping with starting points for the optimizier

fit5 <- sfmfep(formula = y ~ x + (z),
    myPar = c(sigma_u = 0.2, sigma_v = 0.1, beta = 0.5, delta = 0.5),
    estimate = TRUE, bootstrap = TRUE, B = 20,
    method = "firstdiff", panel = sfm.data$producer, data = sfm.data)
summary(fit5)

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