Hi all,
I am working in the field of Financial Services for a reputed MNC and am developing a forecasting model on group level based on Hierarchical time series components constituting of different sub-components and countries (i.e., forecasting for all bottomlevel components and then aggregating them).
But I have two confusions 1) Does using different models for different components and then aggregating them will create problems (in terms of error or causality)?? I mean, will the nature of forecast deviation from actuals change for aggregated component??
And,
2) Is Mape really a standard to measure the accuracy of forecast? if not, then how do I measure the accuracy - with something similar to MAPE or use prdiction interval ?? I feel much more comfortable with the latter.
