Nonlinear regression: finite sample guarantees
share
This paper offers a new approach for study the frequentist properties of the penalized MLE for general nonlinear regression models. The idea of the approach is to relax the nonlinear structural equation by introducing an auxiliary parameter for the regression response and replacing the structural equation with a penalty. This leads to a general semiparametric problem which is studied using the SLS approach from <cit.>. We state sharp bounds on concentration and on the accuracy of the penalized MLE, Fisher and Wilks expansions, evaluate the risk of estimation over smoothness classes, and a number of further results. All the bounds are given in terms of effective dimension and do not involve the ambient dimension of the parameter space.
READ FULL TEXT