|
Consistent Estimation of Minimal Subset Dimension
R. Kohn
Abstract
For a general parametric model we consider the problem of consistently estimating that permissible subset of the parameter space that contains the true parameter point and has smallest dimension. The subset selection is done by means of a model selection criterion of the form used by Akaike [2, 3] and Hannan [5]. Properties of the parameter estimates are discussed and a two step estimation technique is given. The theory is then applied to the logistic regression model.
|
