Application Of Latent Variabal In Different Types Of Data

In biomedical research,multiple indicators of an individual will be observed overdifferent time or different locations, the observations are generrally regarded aslongitudinal data.The indicator tries to describe a potential characteristics ofindividuals from different sides, the potential characteristics of individuals areregarded as latent variables.We try to through the change of response variables touncover the latent trait.When the response variables are continuous and binary, thenwe will take them into exponential family of distributions,as well as latent variableshave independent normal distribution, we will estabish the model between the latentvariables and the response variables, the maximum likelihood estimates of parametersare obtained using a Monte Carlo EM algorithm. When the response variables arecontinuous and ordinal, then we will estabish the model of ordinal will use logisticregression model.In the joint model,response variables are ordered as contious,andthus continuous and ordinal response varibles can be ordered as continous responsevariables, jointly with the latent variable model. The software of SAS PROCNLMIED procedure will cumpute the maximum likelihood estimates of theparameters.