Suchergebnisse

Probability samples are considered. In disproportionate sampling, not all members of the population have an equal chance of being selected, so the results from the various subgroups are weighted to account for the differing probabilities of inclusion. The problem of disproportionate returns due to nonresponse is sometimes treated as equivalent to disproportionate sampling. In the case of disproportionate returns due to nonresponse, however, the situation is different: inclusion is not random & the probability of an individual's inclusion is not known. Another method is to assume that the non-R's are similar to some section of the R's (such as those who made a late return), & to multiply the returns of this section of R's to adjust for nonresponse. Population estimates will then be biased, however, to the extent to which the weighted returns differ from those one would have obtained from the non-R's. The best procedure for avoiding nonresponse bias at the least cost, is to attempt to obtain 100% returns from a random sample of the non-R's. The results of this subsample are then weighted to adjust for the remaining non-R's. 3 Tables. H. Dorian.