The complex sample design requires that the selection probabilities and the field implementation be accounted for in estimating population parameters. The data set contains weights to compensate for differential probabilities of selection and response rates among demographic groups. Analysts should use weights in constructing estimates from the survey and account for the complex sample design in estimating standard errors for survey estimates.

This paper illustrates the use of multilevel statistical modelling of cross-classi®ed data to explore interviewers' in¯uence on survey non-response. The results suggest that the variability in whole household refusal and non-contact rates is due more to the in¯uence of interviewers than to the in¯uence of areas. The results from separate logistic regression models are compared with the results from multinomial models using a polytomous dependent variable (refusals, non-contacts and responses). Using the cross-classi®ed multilevel approach allows us to estimate correlations between refusals and non-contacts, suggesting that interviewers who are good at reducing whole household refusals are also good at reducing whole household non-contacts.

The paper addresses means of generalizing from an experiment based on a nonprobability sample to a population of interest and to subpopulations of interest, where information is available about relevant covariates in the whole population. Using stratification based on propensity score matching with an external populationwide data set, an estimator of the population average treatment effect is constructed. An example is presented in which the applicability of a major education intervention in a non-probability sample of schools in Texas, USA, is assessed for the state as a whole and for its constituent counties. The implications of the results are discussed for two important situations: how to use this methodology to establish where future experiments should be conducted to improve this generalization and how to construct a priori a strategy for experimentation which will maximize both the initial inferential power and the final inferential basis for a series of experiments.

The inclusion of coresident partners enhanced the study by allowing the examination of how intimate, household relationships are related to health trajectories and by augmenting the size of the NSHAP sample size for this and future waves. The uncommon strategy of returning to Wave 1 nonrespondents reduced potential bias by ensuring that to the extent possible the whole of the original sample forms the basis for the field effort. NSHAP Wave 2 achieved its field objectives of consolidating the panel, recruiting their resident spouses or romantic partners, and converting a significant proportion of Wave 1 nonrespondents.