SRMS
Session Slot: 8:30-10:20 Wednesday
Estimated Audience Size: xx-xxx
AudioVisual Request: none dws
Session Title: Analysis of Longitudinal Data from Complex Surveys
Often survey data involve repeated measurements from the
same individual. Rotating Panel Data provides a good example. In such
cases, the assumption of independence for the same individual over
different time points is highly questionable. Building dependence among the
observations through subject-specific random effects is a standard tool
for analyzing such data. The speakers will discuss various aspects of it
both frequentist and Bayesian, and will point out a large
number of applications for complex surveys.
Theme Session: No
Applied Session: ?
Session Organizer: Ghosh, Malay University of Florida
Address:
Phone:
Fax:
Email: ghoshm@stat.ufl.edu
Session Timing: 110 minutes total (Sorry about format):
110 minutes total...please allocate Opening Remarks by Chair - 5 minutes First Speaker - 25 minutes Second Speaker - 25 minutes Third Speaker - 25 minutes Discussant - 20 minutes Floor Discussion - 10 minutes
Session Chair: Ghosh, Malay University of Florida
Address:
Phone:
Fax:
Email:
1. Analysis of Longitudinal Data from Complex Surveys
Rao, J. N. K., Carleton University
Address:
Phone:
Fax:
Email:
Abstract: Semi-parametric marginal generalized linear models are often used for the analysis of longitudinal data. Using such models, Liang and Zeger (1986) obtained standard errors of regression parameter estimates and associated Wald tests, assuming a "working" correlation structure for the repeated measurements on a sample subject. Rotnitzky and Jewell (1990) developed quasi-score tests and Rao Scott adjustments to "working" quasi-score tests. These methods are asymptotically robust to misspecification of the correlation structure, but assume independence of sample subjects which is not satisfied for complex survey data based on stratified multistage sampling. We propose asymptotically valid Wald and quasi-score tests for longitudinal survey data, using the jackknife or the Taylor linearization methods. Alternative tests, based on Rao-Scott adjustments to naive tests that ignore survey design features and on Bonferroni-t, are also developed. These tests are particularly useful when the effective degrees of freedom, usually taken as the total number of sample primary units (or clusters) minus the number of strata, is small.
2. Modelling of Complex Survey Longitudinal Data
Nathan, Gad, Hebrew University of Jeruselum
Address:
Phone:
Fax:
Email:
Pfefferman, Danny, Hebrew University of Jerusalem
Abstract: Longitudinal data consist of repeated measurements on the same sampling units such as households or individuals. Measurements associated with the same unit are naturally intercorrelated, and these correlations need to be accounted for in any model fitting process. Two other important considerations when fitting models to longitudinal data obtained from complex surveys are (1) the causes and pattern of the missing observations, whether implied by the sampling design such as in rotating sample surveys, from leaving the population, or from nonresponse, and (2) the possible effects of the sampling mechanism and in particular, the sample selection probabilities on the model holding for the sample data.In this paper we study plausible models and inference methods that are suitable for fitting short term intercorrelated data that are subjected to sampling effects and (possibly non-ignorable) missing data. In particular, we explore methods of estimating the intercorrelations and incorporating them in subsequent analysis, with special emphasis on multilevel models. These models are in common use for analyzing data with strong intracluster correlations that result from common unobservable factors affecting units in the same cluster. The methods will be illustrated using simulated and real survey data obtained from the Israel Labor Force Survey.
3. Analysis of Longitudinal Binary Data from Complex Surveys: A Hierarchical Bayesian Approach
Lahiri, Partha, University of Nebraska-Lincoln
Address:
Phone:
Fax:
Email: plahiri@unlinfo.unl.edu
Maiti, T., Wu, C., University of Nebraska-Lincoln
Abstract: Longitudinal data analysis has received considerable importance in recent years among survey samplers. In this article, we develop a fully hierarchical model-based methodology to analyse longitudinal binary data that can arise in a complex survey. We carry out our analysis using the Monte Carlo Markov Chain (MCMC) technique. We extend different Bayesian model diagnostic tools to finite population sampling and evaluate them through simulation.
Discussant: Binder, David Statistics Canada
Address:
Phone:
Fax:
Email:
List of speakers who are nonmembers: None