On Friday, August 14, 2015 at 4:56:22 AM UTC-4,
padm...@gmail.com wrote:
> What test (parametric or non parametric) do you suggest to test the significant difference between the predictions made by two different models used in forecasting a given time series?
A reasonable measure of difference is mse(y1-y2)/std(y1)/std(y2). I don't know if this is related to a common measure.