Improving Statistical Efficiency and Testing Robustness of Conjoint Marginal Valuations
研究面板估计量和样本加权如何提高联合分析中边际价值的系数估计效率和置信区间精度，发现随机效应有序probit模型能显著收紧置信区间，而基于人口普查的加权会改变部分边际价值。
Abstract We investigate the effect of panelestimators and sample weighting to improve efficiency of coefficient estimates and confidence intervals on marginal values. Panel estimators are appropriate because most conjoint studies have respondents rate multiple product profiles. Using a random effects ordered probit model increases significance levels on two forest health attribute coefficients and results in substantial tightening of confidence intervals on marginal values. To mitigate the effects of low response rate, we weight the returned surveys using Census data to match the sample to the population. Two of the four marginal values from this weighted ordered probit model are substantially different.