In the class, we show that when there is conditional heteroskedasticity, the GLS model is just a weighted least square model. Suppose there is only serial correlation but there is no conditional heteroskedasticity in the error term, and assume that there the serial correlation only last one period, namely 5,- = pEH + m, where u, is a white noise, is it possible to use the GLS model? If yes, how to do that