Copula Based Stochastic Frontier Quantile Model

Provides estimation procedures for copula-based stochastic frontier quantile models for cross-sectional data. The package implements maximum likelihood estimation of quantile regression models allowing flexible dependence structures between error components through various copula families (e.g., Gaussian and Student-t). It enables estimation of conditional quantile effects, dependence parameters, log-likelihood values, and information criteria (AIC and BIC). The framework combines quantile regression methodology introduced by Koenker and Bassett (1978) with copula theory described in Joe (2014, ISBN:9781466583221). This approach allows modeling heterogeneous effects across quantiles while capturing nonlinear dependence structures between variables.

title: "copulaSQM" output: github_document

copulaSQM

Copula-Based Stochastic Frontier Quantile Model

Description

copulaSQM implements a copula-based stochastic frontier quantile model using the asymmetric Laplace distribution (ALD) and simulated likelihood estimation.

The model extends classical stochastic frontier analysis (SFA) by:

• Estimating conditional quantile frontiers
• Allowing asymmetric error distributions
• Modeling dependence between noise and inefficiency via copulas

This framework is suitable for applied production analysis, efficiency measurement, and heterogeneous performance evaluation.

Model

The production model is:

[ Y_i = X_i \beta + V_i - U_i ]

where:

• ( V_i ) follows an asymmetric Laplace distribution
• ( U_i \ge 0 ) is a one-sided inefficiency term
• Dependence between ( V_i ) and ( U_i ) is modeled using copulas available in VineCopula

Installation

# install.packages("devtools")
devtools::install_github("woraphonyamaka/copulaSQM")

library(copulaSQM)