ENBIS-17 in Naples
9 – 14 September 2017; Naples (Italy)
Abstract submission: 21 November 2016 – 10 May 2017
Keeping Indecision of Respondents in Composite Indicators
12 September 2017, 14:50 – 15:10
- Submitted by
- Domenico Piccolo
- Stefania Capecchi (University of Naples Federico II), Rosaria Simone (University of Naples Federico II), Domenico Piccolo (University of Naples Federico II)
- Composite indicators are valuable tools to quantify the overall assessment of a latent trait provided by multi-item surveys. This feature is important in dealing with massive data set. Several proposals have arisen in the literature in order to meet different objectives and to measure various aspects of the data.
In case evaluations are collected by means of ordinal scales, a mixture model has been developed to take the data generating process into account. In this approach, the final rating is interpreted as the combination of an agreement towards the item under investigation and an inherent indecision accompanying the decision-making process. Such an uncertainty measure accounts for the overall nuisances affecting a fully meditated response. This parametric approach has turned out to be a very effective tool to perform an item by item analysis of questionnaires and comparisons among different items can be effectively yet simply based on similarity among the corresponding agreement and uncertainty parameters.
In order to project CUB models methodology in a multi-item framework, the present contribution aims at integrating it in the process of selection of composite indicators. This task is pursued by exploiting a simple graphical representation of CUB models: thus, composite indicators are mapped in the parameter space after suitable ordinal transformation and CUB estimation. As compared with standard approaches, a distinctive feature of the proposal consists in retaining the level of uncertainty which is generally present in questionnaire analysis.
A case study analyzed via the R package supports the effectiveness of the proposal.
Return to programme