1 comment
  • Official comment

    In regard to the ARIMA model,

    • (Good) The residuals don't have skewness or excess-kurtosis (fat tails), and mean is not different from zero
    • (Good) The residuals are normally distributed.
    • (Bad) The residuals exhibit some serial correlations

    Where we go from here? Experiment with different orders (P,Q) to make the residuals acts like a white-noise.

    I suggest, you difference the input data set, and then run correlogram analysis, to get a good idea of the ARIMA model orders (i.e. P and Q).


    Comment actions Permalink

Please sign in to leave a comment.

Didn't find what you were looking for?

New post