SMABS 2004 Jena University
SMABS 2004 Home Organization About Jena Sponsors Links Imprint / Contact SMABS Home

European Association of Methodology

Department of methodology and evaluation research

Jena University

Contributions: Abstract

The implementation and evaluation of SurveyGlim in LISREL 8.7

Stephen Du Toit Mathilda du Toit Gerhard Mels
Scientific Software Intl
USA

There has been a growing interest in recent years in fitting models to data collected from surveys using complex sample designs. The theory for parameter estimation of generalized linear models (GLIM) is well documented; see e.g. Agresti (1996, 2002) and McCullagh & Nelder (1989).In the literature cited, estimated standard errors of the modelparameters are based on the assumption that the data are from a simple random sample (SRS).

In LISREL we assume that the population from which the sample are obtained can be stratified into H strata. Within each stratum, clusters (primary sample units or PSUs) are drawn and within each stratum-cluster combination, the sampling units (USUs) are drawn with specified design weights. There is also an option to correct for finite populations, provided that the sampling rates or population sizes are available.

The GLIM procedure implemented in LISREL8.7 allows users to select fromthe multinomial, Bernoulli, binomial, Poisson, negative binomial, gamma,Gauss and inverse Gaussian distributions. Various link functions, suchas the log, cumulative logit, cumulative probit, complementary log-logand logit are available.

Use is made of a Taylor linearization procedure to obtain appropriatestandard errors. Results of a simulation study are presented to evaluatethe accuracy of the standard error estimates under various survey designschemes.