Item selection by an extended Latent Class model: An application to nursing homes evaluation

摘要: The evaluation of nursing homes and the assessment quality health care provided to their patients are usually based on administration questionnaires made a large number polytomous items. In applications involving data collected by this type, Latent Class (LC) model represents useful tool for classifying subjects in homogenous groups. paper, we propose an algorithm item selection, which is LC model. proposed aimed at finding smallest subset items provides amount information close that initial set. method sequentially eliminates do not significantly change classification sample with respect full set model, then selection algorithm, may be also used missing responses dealt assuming form latent ignorability. potentialities approach illustrated through application home dataset within ULISSE project, concerns quality-of-life elderly hosted Italian homes. presents several issues, such as very included questionnaire.