Sutradhar, Brajendra C.; Kovacevic, Milorad Analysing ordinal longitudinal survey data: Generalised estimating equations approach. (English) Zbl 1028.62009 Biometrika 87, No. 4, 837-848 (2000). Summary: Longitudinal survey data may comprise ordinal polytomous repeated observations and a set of multidimensional covariates for a large number of individuals. One of the main goals of the longitudinal survey is then to describe the marginal expectation of the ordinal polytomous outcome variable as a function of the covariates while accounting for the structural as well as longitudinal correlations. The structural correlations come from the polytomous nature of the response variable, and the longitudinal correlations from the repetition of the polytomous responses over time.We develop a generalised estimating equations approach based on autocorrelation structure to deal with multivariate polytomous longitudinal survey data, the univariate and bivariate analyses being special cases. The regression estimators are shown to be consistent for the corresponding regression parameters. The methods are illustrated by using the Survey of Labour and Income Dynamics data from Statistics Canada. Cited in 13 Documents MSC: 62D05 Sampling theory, sample surveys 62J02 General nonlinear regression Keywords:contingency table; longitudinal autocorrelation; multinomial logistic distribution; multivariate ordinal categorical data; structural correlation; survey weight PDFBibTeX XMLCite \textit{B. C. Sutradhar} and \textit{M. Kovacevic}, Biometrika 87, No. 4, 837--848 (2000; Zbl 1028.62009) Full Text: DOI