M Guillén1, S Juncà, M Rué, J M Aragay. 1. Dept. d'Econometria, Estadística i Economia Espanyola, Facultat de Ciències Econòmiques i Empresarials, Universitat de Barcelona, Tte. Cor. Valenzuela, 1-11, 08034 Barcelona. guillen@eco.ub.es
Abstract
OBJECTIVE: To illustrate the way in which a sample design can be easily incorporated into the statistical analysis of complex surveys to obtain correct estimates. METHODS: A statistical program (STATA) was used to analyze the Catalan Health Survey (Spain) for the year 1994. RESULTS: If the sample design is taken into account, the estimates are unbiased. If only the weights are taken into account, unbiased point estimates are obtained, but the standard errors tend to be underestimated. CONCLUSIONS: The current availability of statistical programs that allow to incorporate easily the sample design should encourage researchers to better analyze their surveys.
OBJECTIVE: To illustrate the way in which a sample design can be easily incorporated into the statistical analysis of complex surveys to obtain correct estimates. METHODS: A statistical program (STATA) was used to analyze the Catalan Health Survey (Spain) for the year 1994. RESULTS: If the sample design is taken into account, the estimates are unbiased. If only the weights are taken into account, unbiased point estimates are obtained, but the standard errors tend to be underestimated. CONCLUSIONS: The current availability of statistical programs that allow to incorporate easily the sample design should encourage researchers to better analyze their surveys.