Generate data by using the same DGP as that in section 5.3, and implement the first step of PO LS estimation to get the predicted variance varu. Now com-pare several different methods to implement tne second step or weigntea. estima-tion.
First, use regress with the modifier faweight..1/vartil , as in the text.
Second, manually implement this regression by generating the transformed variable tryry/sqrt (van) and regressing try on the similarly constructed variables tr:c2, trx3, and trone, using regress with the noconstant option.
Third, use regress with (pweight=1/varu3, and show that the default standard errors us-ing pweights differ from those using aweights because the pweights default is to compute robust standard errors.