Efficient estimation of maximum likelihood models with multiple fixed-effects: the R package FENmlm
Year: 2018
Type: preprint
Abstract: Fixed-effect models are widely used econometric methods. This paper presents the R package FENmlm which is devoted to the estimation of maximum likelihood (ML) models with any number of fixed-effects. The core of the algorithm, detailed in the paper, is based on a general framework to estimate any ML model with multiple fixed-effects. It also integrates a fixed-point acceleration method to hasten the convergence of the fixed-effect coefficients. The R function offers the user a simple way to estimate any of four different maximum likelihood models: Poisson, Negative Binomial, Gaussian and Logit. Illustrations with real data detail the estimation process as well as the clustering of standard-errors and the various tools to export and manage results from multiple estimations. Simulations show that the algorithm outperforms existing methods in terms of computing time (often by orders of magnitude) or, in the Gaussian case, is on a par with the most efficient ones. Most interestingly, apart from the Gaussian case, the algorithm is revealed to be the only able to estimate models with many fixed-effects on a simple laptop. FENmlm is a free software and distributed under the general public license, as part of the R software project.
Author: Laurent Bergé
Language: English
Topic: Data Analysis with R
Subfield: Statistics and Probability
Field: Mathematics
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2907760633