# Recursive forecast

Dear all,

I have a question regarding recursive forecasting.

Let’s say I have data [1:95] and I want to run a recursive forecast for [84:95], i.e. the first forecasted period is in time 84, and then calculate RMSE for the forecasted period. If I run an estimation with nobs=[84:95], forecast=12, I get a 12x12 matrix in the results. If I get it right, the columns represent k-step prediction and the rows represent a period in which the forecast is calculated.

1. Then, the element [1;1] is a 1-step ahead forecast made in time 83, which means prediction for the period 84, and the other elements of the respective row are predictions for 85, 86,… 95, right?

If so, is the element [12;1] a 1-step ahead prediction for a period 95 made in time 94? And the rest is 96, 97 and so on?

1. Following this logic, if I want to calculate RMSE for the 1-step ahead recursive forecast, I take the first column and compare it to the data, right? That would mean if I want to calculate RMSE up to 12-step ahead forecast, I have (with the nobs=[84:95] and the fact that I only have the data up to the time 95) only one observation of the 12-ahead forecast, so I would need to extend the forecast period to have decent results of 12-ahead forecast?

Thank you very much.

Martin

From the manual

VARIABLE_NAME contains a matrix of the following size: number of time periods for which forecasts are requested using the nobs = [INTEGER1:INTEGER2] option times the number of forecast horizons requested by the forecast option. i.e., the row indicates the period at which the forecast is performed and the column the respective k-step ahead forecast. The starting periods are sorted in ascending order, not in declaration order.

So yes on both counts.