On a Composite Indicator for Observing and Monitoring Scientific Entities: Fair-E

The aim of this article is to present the conceptualization and statistical validation of a unidimensional composite indicator designed to monitor the performance of scientific entities, named the Flexible and Accurate Composite Indicator for Research Entities (Fair-E). A central contribution of the article is its direct response to long-standing criticisms—primarily related to statistical weakness and lack of transparency—commonly directed at composite indicators. The validation procedure addressed the specific characteristics of the dataset through multiple samples and a stress test, including the rotation of factor loadings, the factor extraction method, the use of robust estimators, and the assessment of model fit, ensuring statistical robustness and internal consistency. The absence of weighting was theoretically justified, the use of the Average Variance Extracted (AVE) was dismissed, and unidimensionality was demonstrated both statistically and conceptually. Finally, a Random Forest regression was employed as a robustness and stability check under predictive validation, rather than as evidence of causal inference. The article delivers twelve statistically valid compositions of Fair-E and reports their stability across different sample sizes, allowing application across different types of entities and research fields. The indicator model also supports informed decision-making by combining quantitative analysis, graphical representation, and qualitative interpretation, reinforcing the integration of different analytical perspectives.