Syndicate content

Semi-parametric Models

Parametric modeling provides a convenient inferential setting but is very restrictive. On the other hand, in high-dimensional data, nonparametric models that impose only loose constraints, such as smoothness, require very large sample sizes for convergence of estimators and good inferential performance. Semi-parametric methods allow modeling of some parameters parametrically and others nonparametrically, often improving interpretability and performance of the nonparametric portion without sacrificing efficiency of the parametric portion. This group will explore methods for feature screening or selection prior to modeling as well as methods for joint modeling of the parametric and nonparametric parts of the model.

Sue McDonald's picture

Webex Procedures