Between ICA and SEM

点击翻译按钮获取中文摘要

(1) We have successfully introduced the ICA methodology of confirmatory nature and developed a new estimation method. These and related results were presented in IMPS2003 in Italy and in the annual meeting of the Behaviormetric Society of Japan in 2003.(2) We have proposed a new framework of Nonnormal SEM and applied it to the problem of statistical causal modeling. These and related results were presented in AIC30 in Yokohama and in the annual meeting of the Behaviormetric Society of Japan in 2003.(3) We organized an invited paper session entitled "Nonnormal Structural Equation Modeling" in IMPS2004 in California I myself chaired the session and gave a talk about "Between ICA and SEM."(4) We hosted an international symposium on ICA and SEM at the Nakanoshima Center in October in 2004. We had more than 80 participants.(5) We have studied specific structures on the mixing matrix in ICA in the analysis of risk and protective factors in criminal psychology.(6) In July of 2005, I gave an invited paper entitled "SEM and nonnormality" in the Netherlands in IMPS2005.(7) Using nonnormal information on data, we proposed a solution to the intrinsic problems of normal SEM such as no identifiable models, equivalent models, and saturate models.(8) We suggested to determine causal ordering of several nodes (observable variables) on the basis of ICA and path analysis of SEM.