I'm working with some data.
But it isn't a long series. In fact, it has 47obs.
The data is on levels. So, i worked with the multiplicative model.
I tried different models, and they have the similar Qs and Ms statistics.
The significance of the model's parameters varies.
One issue it's that m8, m10 and m11 are over 1. The Qs statistics are always less than one. This is for all the models estimated.
None of the models have ALL the parameters with statistical significance.
Should i just choose the model with the minor Q and AIC/BIC values?
Can i dismiss the no significance of a parameter and pick the model anyway?
If i can access to a longer series, should i work always with the best model selected by x12 and apply it to the shortest series?