Distance Representations of Consumer Perceptions: Evaluating Appropriateness by Using Diagnostics
研究了如何利用诊断指标（如偏度）来预测空间模型和树模型对消费者感知数据的适用性，发现偏度是最佳预测指标，对营销研究者选择合适模型有参考价值。
To understand consumer perceptions of product/market structures, marketers must choose from a wide variety of spatial and tree models. Because spatial and tree representations in general possess different distance patterns, diagnostic measures calculated from the input data of dissimilarities or similarities should be able to indicate how appropriate a certain type of representation might be for a given set of input data. In this article, the author draws from previous literature on the characteristics of diagnostic measures and representation models to develop some partial hypotheses about how well the measures might indicate the appropriateness (in terms of fit) of different models. Empirical analysis indicates that the skewness diagnostic is clearly the best predictor of the appropriateness of representation models; this finding is consistent across a variety of comparable spatial and tree models. Centrality and the reciprocity-related measure, in conjunction with skewness, are useful for specific types of space–tree pairs. The author uses the U-Method (closely related to jackknifing) of prediction, in conjunction with discriminant analysis models, to show that the diagnostics can predict the relative appropriateness of spaces versus trees with accuracy levels substantially greater than what would be expected by chance.