Estimation and inference for weighted nonlinear least squares regressions are examined for the case in which the regressors are stochastic, rather than fixed, and where erros may be both heteroscedastic and serially correlated. The usual least squares parameter covarience matrix estimator may be...