A Gauss implementation of multivariate sliced inverse regression (Q1381513): Difference between revisions

The Multivariate Sliced Inverse Regression (MSIR) model \[ Y_i=g(X_i'\beta,\varepsilon_i), \;\;\varepsilon_i\perp X_i, \] is considered, where \(Y\in R^q\), \(X_i\in R^p\), are known observations, \(\beta\in R^d\) is an unknown parameter, \(g\) is an unknown \(q\)-dim. vector function. The problem of estimation of \(\beta\) and \(g\) is considered. An iterative algorithm for estimation of \(\beta\) is described and an algorithm using nonparametric regression methods is proposed for estimation of \(g\) in the case when \(g\) is of a special form. A program which implements these algorithms is described.