Department of Mathematical Sciences
|DATE:||Thursday, Sept. 5, 2019|
|TIME:||1:15pm – 2:15pm|
|SPEAKER:||Mengyu Chen, Binghamton University|
|TITLE:||Weighted least squares estimation: An empirical likelihood approach|
Abstract: For the heteroscedastic linear model, a possible estimator of the regression parameter theta is the weighted least squares estimator. However, the best weighted least squares estimator relies on the conditional variance function, which is usually unknown.
The usual method is constructing an estimator of the variance function. Instead, we can use a maximum empirical likelihood estimator which is based on an increasing number of estimated constrains and avoids estimating the variance function.