Using a Generalized Linear Mixed Model to Analyze Dichotomous Choice Contingent Valuation Data
研究了二分选择条件估值中支付意愿数据的过度离散问题，提出用广义线性混合模型来减少参数估计偏差和显著性高估，并用已发表数据举例说明。
Willingness-to-pay (WTP) responses from dichotomous choice contingent valuation studies are often modeled using logistic regression, from which estimates of mean or median WTP are calculated. However, a great many factors influence an individual's WTP, some of which may be unobserved. Hence, the regression model may have inadequate explanatory power, and parameter estimates may be biased and their significance overestimated. The effects of this overdispersion are examined within a generalized linear mixed modeling framework, and an example given using published data from Cooper and Loomis (1992).