Model averaging with averaging covariance matrix

- We use an average estimator which depends on all candidate models to estimate the covariance matrix.
- We choose weight vectors in the model average estimators of coefficients and covariance matrix simultaneously by minimizing the weight choice criterion.
- We prove the asymptotic optimality.
- Simulation experiments show that the proposed model averaging method is superior to its competitors.

This article studies optimal model averaging for linear models with heteroscedasticity. We choose weights by minimizing Mallows-type criterion. Because the covariance matrix of random error in the criterion is unknown, an averaging estimator of covariance matrix is plugged into the criterion. The resulting model averaging estimator is proved to be asymptotically optimal under some regularity conditions. Simulation experiments show that the proposed model averaging method is superior to its competitors.