Jensen E
Demography. 1985 Aug;22(3):445-54.
An unresolved question in the analysis of survey data relating to fertility attitudes and beliefs is how non-numeric responses to questions on ideal family size should be treated. This paper demonstrates that simply dropping "up to God" responses will bias regression results. An unbiased estimator is presented which explicitly models the way in which observations are selected into the sample. The estimator is then employed on Guatemalan and Indian data. No support for the notion that women who answer "up to God" are women who would have given relatively large numeric answers is found in these two samples.
在分析与生育态度和观念相关的调查数据时，一个尚未解决的问题是，对于理想家庭规模问题的非数字回答应如何处理。本文表明，简单地剔除“听天由命”的回答会使回归结果产生偏差。本文提出了一种无偏估计量，该估计量明确地对样本中观测值的选择方式进行了建模。然后将该估计量应用于危地马拉和印度的数据。在这两个样本中，未发现支持以下观点的证据：回答“听天由命”的女性是那些原本会给出相对较大数字回答的女性。
