Scientific article
French

Stochastic variable selection strategies for zero-inflated models

Published in

Statistical modelling, vol. 18, no. 1, p. 3-23

Publication date

2018

Abstract

When count data exhibit excess zero, that is more zero counts than a simpler parametric distribution can model, the zero-inflated Poisson (ZIP) or zeroinflated negative binomial (ZINB) models are often used. Variable selection for these models is even more challenging than for other regression situations because the availability of p covariates implies 4p possible models. We adapt to zero-inflated models an approach for variable selection that avoids the screening of all possible models. This approach is based on a stochastic search through the space of all possible models, which generates a chain of interesting models. As an additional novelty, we propose three ways of extracting information from this rich chain and we compare them in two simulation studies, where we also contrast our approach with regularization (penalized) techniques available in the literature. The analysis of a typical dataset that has motivated our research is also presented, before concluding with some recommendations.

Keywords

  • Excess zero
  • ZI model
  • Hurdle model
  • Variable selection
  • Stochastic search

Citation (ISO format)

CANTONI, Eva, AUDA, Marie. Stochastic variable selection strategies for zero-inflated models. In: Statistical modelling, 2018, vol. 18, n° 1, p. 3–23. doi: 10.1177/1471082X17711068

Main files (1)

Article (Accepted version)

Identifiers

Journal ISSN

1471-082X
593views
0downloads

Technical informations

Creation

08/07/2017 23:22:00

First validation

08/07/2017 23:22:00

Update

15/03/2023 01:54:44

Status update

15/03/2023 01:54:44

Last indexation

29/10/2024 11:21:24
All rights reserved by Archive ouverte UNIGE and the University of Geneva
OpenAIRE Indexed & Validated