|
Abstract:
|
Sample surveys can have complex sample designs with multistage-cluster sampling, stratification, and differential selection probabilities from a target finite population. To make inferences to population parameters, sample weighting and appropriate estimation of variance-covariances in the analysis is usually required. Because of desirable parameter invariance and small sample properties of generalized score tests for simple random samples, we propose quasi-generalized score hypothesis tests of functions of population parameters that allow for general parameter transformations and constraints for complex samples. We obtain score vectors from sample weighed estimating equations and estimate variance-covariance for the scores by Taylor linearization methods that account for the complex sample design. We illustrate the proposed tests with two applications, testing a coefficient of variation and goodness-of-fit for logistic regression. We investigate the finite sample type I error rates of the proposed methods by Monte Carlo simulation studies and further illustrate the methods with application to data from the continuous National Health and Nutrition Examination Survey.
|
