Jay G M, Liang J, Liu X, Sugisawa H
Survey Research Center, University of Michigan.
J Gerontol. 1993 May;48(3):S143-52.
Data from a nationwide sample of older adults in Japan were analyzed to determine patterns of unit nonresponse. Maximum likelihood logistic regression analyses identified age, sex, geographic region, and urbanicity as significant predictors of response status. The probability of nonresponse was higher for men, persons living in certain regions, and those in large urban areas. Age had a curvilinear relationship with response status; the probability of nonresponse decreased from age 60 to age 70, and then increased again with the highest rates at the oldest ages. When nonrespondents were disaggregated into three different groups based on reason for nonparticipation (i.e., refusals, not-at-homes, other reasons), variation in the results was found. For example, while men were more likely than women to refuse an interview or be absent from home during the survey period, they were no more likely to not participate for other reasons. Use of different response rate calculations and a substitution sampling procedure were not found to seriously affect study results.
对来自日本全国老年人样本的数据进行了分析，以确定单位无应答模式。最大似然逻辑回归分析确定年龄、性别、地理区域和城市化程度是应答状态的重要预测因素。男性、居住在某些地区的人和大城市地区的人无应答的概率更高。年龄与应答状态呈曲线关系；无应答概率从60岁到70岁下降，然后在最年长的年龄段再次上升且上升速率最高。当根据不参与的原因将无应答者分为三个不同组（即拒绝、不在家、其他原因）时，发现结果存在差异。例如，虽然男性比女性更有可能拒绝采访或在调查期间不在家，但他们因其他原因不参与的可能性并不更高。未发现使用不同的应答率计算方法和替代抽样程序会严重影响研究结果。
