Variables in psychological research are usually latent, erroneous or unobservable. Traditional statistical analyses do not consider the erroneous property of the observed variables and the structure of the latent concepts simultaneously. Recent developments in latent variable models have been trying to handle this problem. Since LISREL program has increasingly become popular, structural equation modelling with LISREL are introduced briefly and then served as illustrating examples for four methodological issues, which include: (a) decision between latent and non-latent variable models can make a difference on results and conclusions; (b) a model is never confirmed but only disconfirmed; (c) a probabilistic conception of causality is needed; and (d) all current latent variable models treat the latent variable as the cause of its indicators.
[ Chinese version | Index | Journal of Education and Psychology | Other Journals | Faculty Home Page | Enquiry ]
Mail any comments and suggestions to
en@fed.cuhk.edu.hk.