 |
The 130th Annual Meeting of APHA |
|
The 130th Annual Meeting of APHA |
Chin-Lin Tseng, DrPH, National Center for Children in Poverty, Columbia University, 154 Haven Avenue, 3rd Floor, New York, NY 10032, 732-5486171, chinlintseng@yahoo.com
Public health Studies involving neighborhood/area effects on health have dramatically increased in the past few years. Many studies focus on individual health outcomes but incorporate both effects of individual (micro-level) and neighborhood (macro-level/context) variables. The multi-level (random effects) approach has been used most often to handle such data. This study aims to demonstrate that the generalized estimating equation (GEE) approach may be a preferable method to analyze such data of clustered nature by analyzing some dataset to investigate the relationship between individual unemployment and depressive symptom scores, and whether this relationship is affected by the economic context. The statistical analyses are complicated because the data are correlated due to repeated measurements for individuals and the complex survey design itself. GEE and random effects (RE) models using SAS statistical software are compared.
Regression estimates from GEE and RE linear models are almost identical. Their standard errors are less comparable. Computation time for both models are comparable. The personal unemployment effect generally is not affected by the economic context.
The GEE approach may be preferred if the focus of a study is effect estimates of independent variables; it does not need specific assumption of the joint distribution of the observations and has the advantages of yielding consistent estimates of regression coefficients as well as their variances even when the correlation structure of the data is mis-specified. Additionally, When there are tremendous numbers of levels in a cluster, the GEE approach may take less computation time and consume less computer memory.
Learning Objectives:
Keywords: Statistics, Mental Health
Presenting author's disclosure statement:
I do not have any significant financial interest/arrangement or affiliation with any organization/institution whose products or services are being discussed in this session.