Diagnostic Testing in Missing Data Models

Consideration is given to a class of missing data models which has attracted a good deal of attention in labor economics and in the analysis of various social experiments. Focus is placed on the testing of the assumptions of homoscedasticity and lognormality in these models. The results were similar...

Full description

Saved in:
Bibliographic Details
Published in:International economic review (Philadelphia) 1983-10, Vol.24 (3), p.537-546
Main Authors: Poirier, Dale J., Ruud, Paul A.
Format: Article
Language:English
Subjects:
Applications
Consistent estimators
Data models
Econometrics
Economic models
Economic theory
Estimating techniques
Estimators
Labor economics
Lagrange multiplier
Maximum likelihood estimation
Missing data
Modeling
Null hypothesis
Random variables
Statistical analysis
Truncation
Citations: Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Consideration is given to a class of missing data models which has attracted a good deal of attention in labor economics and in the analysis of various social experiments. Focus is placed on the testing of the assumptions of homoscedasticity and lognormality in these models. The results were similar to the analyses of Hurd (1979) in the case of heteroscedasticity and of Goldberger (1980) in the case of nonnormality. Violation of either of these assumptions resulted in inconsistency of the usual estimators proposed for the missing data model considered. The Lagrange multiplier tests that are proposed are asymptotically locally most powerful with regard to small perturbations from the parametric specification of the null hypotheses. It is proposed that they will be reasonably powerful with respect to other deviations from the model's assumptions. The tests are illustrated with an application to the estimation of wage equations for young males.
ISSN:0020-6598
1468-2354
DOI:10.2307/2648784