Gaining Longitudinal Insights from Repeated Cross-Sectional Survey Data: With Simulation-Based Validation and Application

Li, Qingfeng. (2013). Gaining Longitudinal Insights from Repeated Cross-Sectional Survey Data: With Simulation-Based Validation and Application. Master's thesis / Doctoral dissertation, The Johns Hopkins University.

Citation:

Li, Qingfeng. (2013). Gaining Longitudinal Insights from Repeated Cross-Sectional Survey Data: With Simulation-Based Validation and Application. Master's thesis / Doctoral dissertation, The Johns Hopkins University.

Citation Export File (RIS Format):
Octet Stream icon

1955.ris — Octet Stream, 2 kB (3,058 bytes)

Abstract:

Background: Cross-sectional data have been extensively used in public health and social science in general due to their wide availability and simplicity of model estimation and interpretation. But they have a few limitations. This thesis is concerned with two of the major limitations: inability to eliminate bias caused by unobserved heterogeneity; inability to include lagged independent variables. Models for longitudinal data can overcome these limitations under reasonable assumptions, thanks to the availability of repeated measures of the same analysis unit. Unfortunately for public health researchers, longitudinal data are infrequently collected on large populations in most countries. Method: This study has reviewed an approach proposed by Deaton to estimate models for pseudo panels constructed from repeated cross section (RCS) data. We use simulated data and real longitudinal data to validate the pseudo panel models. Then the approach was applied to the investigation of life course impacts of early maternal risk factors on subsequent reproductive and developmental outcomes using data from Demographic and Health Surveys (DHS). Result: The simulation results validated the statistical properties of the pseudo panel estimators. We found that the biases are often small even when the assumptions to ensure the unbiasedness and consistency of the estimators are violated. It implies that pseudo panel estimator should be more frequently used in public health studies. Conclusion: As far as we know, this study is among the first to introduce the pseudo panel approach to public health studies, which reply on quality data and appropriate models. The application study in chapter 5 contributes to our knowledge of life course impacts of risk factors at daughters' birth on their adult developmental and reproductive outcomes based on a pseudo-cohort analysis. Interventions that target eliminating risk factors at birth, such as promotion of adequate birth spacing, can prevent adverse birth outcomes in the long run and in an enduring manner.

URL:

Keyword(s):

Notes:

Reference Type:
THES

Secondary Title:

Series Title:

Author(s):
Li, Qingfeng

Editor(s):
M., Bishai David

Series Author(s):
Tsui, Amy O.

Year Published:
2013

Date Published:

Journal Name:

Volume Number:
3578791

Issue Number:

Start Page Number:
210

End Page Number:

Edition:

Series Volume:

Publisher:
The Johns Hopkins University

City of Publication:
Ann Arbor

ISSN/ISBN:
9781303732508

DOI:

PMCID:

NIHMSID:

Reference ID:
1955

Miscellaneous: