Description: A model for two-stage cluster sampling when sample cluster sizes are unknown is used to derive an optimal (model-based) estimator for the population total and to determine robust sampling strategies. In an empirical study using 1970 and 1980 census data for Los Angeles and surrounding counties, comparisons are made between the model based estimator and conventional estimators. The results favor the new estimator over those derived from randomization theory. In addition, the empirical study shows that the robust sampling strategies suggested by the theory can reduce biases, improve efficiency, and decrease the frequencies of large errors. 7 refs.
Date: January 1, 1985
Creator: Kelly, E.J. & Cumberland, W.G.
Item Type: Refine your search to only Article
Partner: UNT Libraries Government Documents Department