Misspecification effects in the analysis of panel data
Misspecification effects in the analysis of panel data
Misspecification effects (meffs) measure the effect on the sampling variance of an estimator of incorrect specification of both the sampling scheme and the model considered. We assess the effect of various features of complex sampling schemes on the inferences drawn from models for panel data using meffs. Many longitudinal social survey designs employ multistage sampling, leading to some clustering, which tends to lead to meffs greater than unity. An empirical study using data from the British Household Panel Survey is conducted, and a simulation study is performed. Our results suggest that clustering impacts are stronger for longitudinal studies than for cross-sectional studies, and that meffs for the regression coefficients increase with the number of waves analysed. Hence, estimated standard errors in the analysis of panel data can be misleading if any clustering is ignored
487-505
Vieira, Marcel de Toledo
d78eb443-9a6e-400d-a534-5e9a6c56ddf0
Smith, Peter
961a01a3-bf4c-43ca-9599-5be4fd5d3940
Salgueiro, Maria de Fátima
b6ea9210-a0a1-4507-ab7a-f2f64f62c9a4
May 2016
Vieira, Marcel de Toledo
d78eb443-9a6e-400d-a534-5e9a6c56ddf0
Smith, Peter
961a01a3-bf4c-43ca-9599-5be4fd5d3940
Salgueiro, Maria de Fátima
b6ea9210-a0a1-4507-ab7a-f2f64f62c9a4
Vieira, Marcel de Toledo, Smith, Peter and Salgueiro, Maria de Fátima
(2016)
Misspecification effects in the analysis of panel data.
Journal of Official Statistics, 32 (2), .
(doi:10.1515/jos-2016-0025).
Abstract
Misspecification effects (meffs) measure the effect on the sampling variance of an estimator of incorrect specification of both the sampling scheme and the model considered. We assess the effect of various features of complex sampling schemes on the inferences drawn from models for panel data using meffs. Many longitudinal social survey designs employ multistage sampling, leading to some clustering, which tends to lead to meffs greater than unity. An empirical study using data from the British Household Panel Survey is conducted, and a simulation study is performed. Our results suggest that clustering impacts are stronger for longitudinal studies than for cross-sectional studies, and that meffs for the regression coefficients increase with the number of waves analysed. Hence, estimated standard errors in the analysis of panel data can be misleading if any clustering is ignored
Other
jos-2016-0025.pdf_t_ac=j$002fjos.2016.32.issue-2$002fjos-2016-0025$002fjos-2016-0025.xml
- Version of Record
More information
Accepted/In Press date: 1 January 2016
Published date: May 2016
Organisations:
Statistical Sciences Research Institute
Identifiers
Local EPrints ID: 397233
URI: http://eprints.soton.ac.uk/id/eprint/397233
ISSN: 0282-423X
PURE UUID: 81f26f94-4a76-464c-89e2-88b4c14c8740
Catalogue record
Date deposited: 24 Jun 2016 11:25
Last modified: 15 Mar 2024 02:43
Export record
Altmetrics
Contributors
Author:
Marcel de Toledo Vieira
Author:
Maria de Fátima Salgueiro
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics