Assessment Method for Weighting and Aggregation in Constructing Composite Indicators of Mixed Data

Abstract

Composite indicators are often encountered in various studies, especially in the social sector. Composite indicators are constructed from several steps such as weighting and aggregation. The classical weighting method such as weighting based on factor analysis and regression analysis cannot handle a mixture of numeric and categoric variables. Furthermore, using a dependent variable as the estimator in weighting based on regression analysis is sometimes manipulated by respondents. An approach to address this problem uses the weighting method based on factor analysis of mixed data. The classical aggregation method such as linear additive aggregation cannot handle a mixture of compensatory and non-compensatory numeric variables. Therefore, to address this problem, a geometric aggregation was used. The case study constructed the five models of household welfare status of Dramaga village, Bogor regency that used the combination of weighting method based on multiple correspondence analysis and factor analysis of mixed data and linear and geometric aggregation. The five models are compared. The best model was model using the weighting method based on factor analysis of mixed data and the geometric aggregation for the numeric variables and the linear aggregation for the categoric variables.