Abstract:
|
Choosing suitable variable transformation is a widely discussed problem in multivariate linear models. We consider the problem of choosing a suitable transformation of a multivariate response variable using a basis spline method. Based on the appropriate transformation method, we use then apply the envelope method to reduce the dimensions of both predictor and response. We demonstrate how to find the transformation such that the linearity between predictor and response is maximized, and demonstrate our proposed approach on extensive numerical experiments.
|