We develop a Bayesian method for heterogeneous variable selection in both linear and nonlinear panel data models, where variable selection takes place at the individual level and non-zero parameters are allowed to differ across individuals. Each individual-specific parameter is either zero or comes from a Dirichlet process mixture of multivariate normals. For inference, we develop an efficient MCMC sampler. In a Monte Carlo study, we show that our method accurately captures complex continuous cross-sectional heterogeneity and individual-specific variable selection, features standard approaches fail to capture jointly. An application on data from a discrete choice experiment on food choices shows that accounting for heterogeneous variable selection and non-normal continuous heterogeneity uncovers substantial variable non-attendance and an improved out-of-sample fit

Castelein, Anoek; Koobs, Stan; Fok, Dennis; Paap, Richard

We develop a Bayesian method for heterogeneous variable selection in both linear and nonlinear panel data models, where variable selection takes place at the individual level and non-zero parameters are allowed to differ across individuals. Each individual-specific parameter is either zero or comes from a Dirichlet process mixture of multivariate normals. For inference, we develop an efficient MCMC sampler. In a Monte Carlo study, we show that our method accurately captures complex continuous cross-sectional heterogeneity and individual-specific variable selection, features standard approaches fail to capture jointly. An application on data from a discrete choice experiment on food choices shows that accounting for heterogeneous variable selection and non-normal continuous heterogeneity uncovers substantial variable non-attendance and an improved out-of-sample fit

Anoek Castelein, Stan Koobs, Dennis Fok and Richard Paap
Additional contact information
Anoek Castelein: Erasmus University Rotterdam
Stan Koobs: Erasmus University Rotterdam

No 20-061/III, Tinbergen Institute Discussion Papers from Tinbergen Institute

Keywords: Attribute non-attendance; Bayesian variable selection; Dirichlet process; Panel data; Unobserved heterogeneity (search for similar items in EconPapers)
JEL-codes: C11 C23 C25 (search for similar items in EconPapers)
Date: 2020-09-22, Revised 2026-03-05
New Economics Papers: this item is included in nep-dcm, nep-ecm and nep-ore
References: View references in EconPapers View complete reference list from CitEc
Citations:

Downloads: (external link)
https://papers.tinbergen.nl/20061.pdf (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:tin:wpaper:20200061

Access Statistics for this paper

More papers in Tinbergen Institute Discussion Papers from Tinbergen Institute Contact information at EDIRC.
Bibliographic data for series maintained by Tinbergen Office +31 (0)10-4088900 ().