Application of ICA for Multiple Regression

Independent Component Analysis(ICA) is a widely applicable modern method for finding underlying factors or components from multivariate statistical data. In this paper, we consider the application of ICA for multiple regression. When multicollinearity is present in a set of explanatory variables, the ordinary least squares estimate of the individual regression coefficient tend to be unstable and can lead to erroneous inferences. Numerical study shows we can have stable parameter estimates based on ICA regression when multicollinearity exist.