[Gretl-users] Estimation of ARIMA

Q1: I input monthly data from 1980M1 to 1990M12. When estimating AR(3), I set the sample range from 1980M6 to 1990M12. If data from 1980M1 to 1980M5 would be included to estimate AR(3), when I set the sample range from 1980M10 to 1990M12, data from 1980M1 to 1980M9would be included to estimate AR(3), is it right ? Does that mean no matter what range I set, data included to estimate AR(3) is always from 1980M1 to 1990M12 ? ( outcome of different data range are different) Q2: When comparing MAPE of AR(1) and AR(2), their MAPE value is gained from the same forecasting range (1989M1 to 1990M12). If I can compare their MAPE value when using observations of AR(1) and AR(2) are different , Is it would be much more appropriate to set the using observations of AR(1) and AR(2) the same? What difference between these two kinds of situations? Under different periods of using observations, If MAPE value (both gain from1989M1 to 1990M12) of AR(1) is lower than AR(2), Can I say forecasting ability of AR(1) is better than AR(2)? Are their basis of comparison different?