Suchergebnisse

Data from a fully enumerated general pop sample were used estimate the magnitude of nonresponse bias & the effect of personal follow-up in a mail survey. An area-probability sample of 4,735 addresses in Alameda County, Calif yielded 8,267 adults. Of these, 8,083 were represented in an initial 'enumeration' survey in which demographic & SE data were gathered by interviewers. These data are the 'parameters' with which subsequent 1 returns were compared. Questionaires were left for each adult in the household at the time of enumeration, with the request that they be completed & mailed back within a week. 2 further attempts-by letter & telegram-were made to obtain responses. Finally, interviewers visited those who had not yet responded, to retrieve completed questionnaire's or help with their completion. Of the 8,083 enumerated adults, 70% responded by mail, 16% were retrieved by interviewers, & 14% did not respond. The nonR's were so small a proportion of the enumeration that their omission had very little effect on the sample estimates for the 13 characteristics considered; they were older, & more likely to be white, M, skilled, & to live in small households without children, but the largest cliff between the questionaire sample & the total/sum enumerated sample was only 1.4% points. Comparisons between mail R's & those whose questionaires were retrieved by interviewers showed that the 2 response groups were very similar. Personal contacts did improve estimates of racial & some related SE distributions, but at a per-R cost of over twice that of the mail response. The personally-retrieved questionnaire's raised estimates of some categories of people easy-tofind-at-home above the actual distribution. These findings indicate that surveys conducted entirely by mail can achieve a good representation of the sample universe, given the high overall return rate that characterized this survey. AA.

A study of reliability or stability of response was conducted with a probability sample of 1,530 Alameda County, Calif, residents. 2 identical questionaires were completed by R's within approximately 2 weeks. 3 out of 4 R's answered at least 90% of the items identically; only 1 in 50 did so for as few as 70-79% of the items. Within this rather narrow range, reliability varied with educ, age, & length of time between questionaires. Questions dealing with objective fact were most reliable, those dealing with att's & states of mind, least reliable. The small number of responses indicating poor health, marital dissatisfaction or a psychol'al problem were more likely to change for the better than responses indicating no such problem were to change for the worse. AA.