It does this by evaluating the prediction errors of the two models about a particular interval. The test checks the null hypothesis the two products contain the identical overall performance on normal, against the alternative that they don't. If the examination statistic exceeds a significant value, we reject the null hypothesis, indicating that the real difference in the forecast accuracy is statistically considerable.

We may also explicitly set the Home windows, seasonal_deg, and iterate parameter explicitly. We will get a worse healthy but This can be just an illustration of more info how you can move these parameters into the MSTL class.

, is undoubtedly an extension in the Gaussian random wander course of action, through which, at each time, we may perhaps take a Gaussian move by using a probability of p or stay in exactly the same condition having a likelihood of one ??p

windows - The lengths of every seasonal smoother with regard to each period of time. If they're significant then the seasonal element will clearly show significantly less variability eventually. Need to be odd. If None a set of default values determined by experiments in the first paper [one] are utilized.