Gini heterogeneity index for detecting uncertainty in ordinal data surveys

摘要

In sample surveys where people are asked to give responses to items by means of ordinal scales it is common to register an inherent indecision generated by both objective and subjective causes. Thus, an effective statistical analysis should take this component into account to avoid bias in estimation, interpretation and prediction. In this paper, we show that the heterogeneity index proposed by Gini and its variants are effective measures to detect such an uncertainty. The relationships of Gini heterogeneity index with the parameter of a mixture model are discussed and exploited as a tool for the exploratory selection of covariates. Then, a real case study is considered to check for the effectiveness of these proposals. Some concluding remarks end the paper.