IMsModelPeriod.AutoPeriod

When forming of sample and forecasting periods using actual data, the model specific is taken into account:

• Actual data of input series used to form the sample period is taken from input variable for the following models: linear regression (OLS, instrumental variables estimation), binary choice model, non-linear regression, trend with selection of functional dependency, all smoothing methods, ARIMA, error correction model, cumulative method, missing data treatment.

• For the Collapse (series calculation), Collapse (pointwise calculation) and Interpolation methods the actual data for the input series to form the sample period is taken from the input variable, by the source frequency.

• If the model does not require the input variable (for example, in a determinate equation and custom methods), sample period start is equal to the minimum date corresponding to the first non-empty values of model operands, sample and forecasting period end is formed according to the maximum date corresponding to the last non-empty values of model operands.

• For the Aggregation model, limits of the sample and forecasting periods are based on the output variable (if the variable does not have slices), or based on the values of the variable by the slices (if the variable has slices).

• Sample start and end dates and forecast end dates are set equal to the calendar start of the modeling container for the models that allow for the use of empty series as the input or input variable and factors.

• On calculating the forecasting period end date equal to last actual date of model factors, the data of output variable is taken into account. If the forecasting period end date is less than the forecasting period start date, the sample period end date is specified as the forecasting period end date.

• For Aggregation (Extended) and Aggregation (Basic) models, the conditions of setting sample and forecasting periods by actual data of variables are not used. Other conditions are implemented in the same way as the ordinary models.

• For non-linear equation systems, input-output balance model, vector error correction model, models on panel data, and the Value-at-Risk model, no conditions of setting sample and forecasting periods by actual data of variables are used. Other conditions are implemented in the same way as the ordinary models.

• If the model has through dimension, the calculation period for it is formed by the first element of the dimension.

On forming the periods using start/end of the sample/forecasting dates the settings of the calendar frequency are taken into account: offset of the period start relative to its start/end.

If the current date is used on forming the period, the end dates of the sample and forecasting are offset to the period end. It is relevant for the monthly frequency and higher.
For example, sample period end is April 2012 (IdentEndAsNow), and the calculation end date is equal to the sample period end date (ForecastEndByIdentEnd). Thus, on calculating the model the calculation end date is 30.04.2012.