Skip to content

ParallelGSReg/JuliaCon2019

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 

Repository files navigation

JuliaCon 2019 Proceedings

GlobalSearchRegression.jl: Building bridges between Machine Learning and Econometrics in Fat-Data scenarios

Authors

Panigo Demian (CONICET, UNLP, UNDAV), Pablo, Gluzmann (CONICET, UNLP), Esteban Mocskos (CONICET, UBA), Adan Mauri Ungaro (UNLP), Valentin Ungaro (UNLP) and Nicolás Monzón (UNLP, UNDAV).

Abstract

The aim of this paper is twofold. The first one is to describe a novel research-project designed for building bridges between machine learning and econometric worlds. The second one is to introduce the main characteristics and comparative performance of the first Julia-native all-subset regression algorithm included in GlobalSearchRegression.jl (v.1.0.5). As other available alternatives, this algorithm allows researchers to obtain the best model specification among all possible covariate combinations - in terms of user defined information criteria-, but up to 3165 and 197 times faster than STATA and R alternatives, respectively.

Credits

The GSReg module, which perform regression analysis, was written primarily by Demian Panigo, Valentín Mari and Adán Mauri Ungaro. The GlobalSearchRegression.jl module was inpired by GSReg for Stata, written by Pablo Gluzmann and Demian Panigo.

About

JuliaCon 2019 archives for proceedings paper

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published