This paper is concerned with time-series forecasting based on the line
ar regression model in the presence of AR(1) disturbances. The standar
d approach is to estimate the AR(1) parameter, rho, and then construct
forecasts assuming the estimated value is the true value. We introduc
e a new approach which can be viewed as a weighted average of predicti
ons assuming different values of rho. The weights are proportional to
the marginal likelihood of rho. A Monte Carlo experiment was conducted
to compare the new method with five more conventional predictors. Its
results suggest that the new approach has a distinct edge over existi
ng procedures.