A Weighted Composite Likelihood Approach to Inference from Clustered Survey Data Under a Two-Level Model

Two-level models are widely used for analysing clustered survey data with the design structure matching model hierarchy. Hypothesis testing on parameters is studied, using a weighted composite likelihood approach that takes account of features. In particular, asymptotic normality estimators established. Using this result, distributions generalised score test statistic and ratio type statistic, under null hypothesis parameters, Results limited simulation study finite sample performance proposed tests reported.