Evaluation of Information Retrieval Systems Using Structural Equation\n Modelling

The interpretation of the experimental data collected by testing systems\nacross input datasets and model parameters is of strategic importance for\nsystem design and implementation. In particular, finding relationships between\nvariables and detecting the latent variables affecting retrieval performance\ncan provide designers, engineers and experimenters with useful if not necessary\ninformation about how a system is performing. This paper discusses the use of\nStructural Equation Modelling (SEM) in providing an in-depth explanation of\nevaluation results and an explanation of failures and successes of a system; in\nparticular, we focus on the case of Information Retrieval.\n