Publication Type

Journal Article

Version

acceptedVersion

Publication Date

2-2022

Abstract

In missing data problems, missing not at random is difficult to handle since the response probability or propensity score is confounded with the outcome data model in the likelihood. Existing works often assume the propensity score is known up to a finite dimensional parameter. We relax this assumption and consider an unspecified single index model for the propensity score. A pseudo-likelihood based on the complete data is constructed by profiling out a synthetic distribution function that involves the unknown propensity score. The pseudo-likelihood gives asymptotically normal estimates. Simulations show the method compares favorably with existing methods.

Keywords

Missing not at random, Nonignorable missing, Pseudo-conditional likelihood, Single index model, Synthetic distribution

Discipline

Econometrics | Economics

Research Areas

Econometrics

Publication

Journal of Business and Economic Statistics

Volume

40

Issue

2

First Page

715

Last Page

717

ISSN

0735-0015

Identifier

10.1080/07350015.2020.1860065

Publisher

Taylor & Francis

Embargo Period

4-29-2021

Copyright Owner and License

Authors

Additional URL

https://doi.org/10.1080/07350015.2020.1860065

Included in

Econometrics Commons

Share

COinS