Modelo lineal general mixtoAlternativa a los modelos de análisis clásicos

Abstract

In function of the reach of the inferences to carry out and of the distribution of the errors associated with the observation units, are obtained four of the models more used in education and psychology, that is: the general lineal model, the generalized lineal, the general linear mixed and the generalized mixed. When the dependent variable has a metric structure, the classic model it is appropriate to relate this with one or more variable of fixed effects. When the dependent variable is categorical, it is inappropriate to assume that the errors are normally; in this case, the natural solution provides the generalized lineal model. Unfortunately, the benefits uf this approach are available only in fixed-effects models. When the dependent variable has a metric structure and multiple sources of random variation, the mixed mude! al!ows a mix of fixed and random effects. None of the previous models is appropriate when the dependent variable is categorical and sorne factors are viewed as random. The generalized mixed model so!ves this problem assuming that the lineal predictor is a function both of the parameters of fixed effects and random.