One Case of Reduction of Nonlinear Regression to Linear


Z. Ivković


An effective procedure to find nonlinear regression $X^*$ of $X=X(\xi_1,\ldots,\xi_M)$ on $\eta_1,\ldots,\eta_N$ in terms of the linear regression coefficients of $\xi_k$ on $\eta_1,\ldots,\eta_N$ is proposed. The variables $\xi_1,\ldots,\xi_M,\eta_1,\ldots,\eta_N$ are Gaussian. Some applications to $N$-ple Markov process are considered too.