Modelli di equazioni strutturali con variabili dicotomiche e non normali: un'applicazione per l'analisi della attività fisica regolare negli anziani / Structural equation modelling with dichotomous and non-normal variables: An application regarding regular practice of physical activity in older adults

In the present study the impact of attitudes, subjective norm, perceived behavioral control and self-efficacy on the intention to regularly practice physical activity has been analyzed. 1095 (65-90 years) old subjects enrolled in the programs of the "Unione Italiana Sport per Tutti" have been considered. Before analyzing the data, variables with a strong violation from normality have been dichotomized and data have been analyzed through the software MPLUS (Muthen & Muthen, 1998). This software allows to consider a general structural equation model with dichotomous, ordered categorical and a combination of continuous and categorical latent variables indicators. Results indicate that the model considered, closely fits the data (21,N= 1085) = 22.23, p =.39; CFI = 1; NNFI = 1; RMSEA = .01]. To compare different approaches to non-normalized data, fit indexes of the model were also assessed with three different methods for parameters esteem: the weighted least squares (WLS) and maximum likelihood (ML) methods, and the "robust" maximum likelihood method (MLMV).