Extension of time series models with non-stationary and uncertain variance serves to address their shortcomings when it comes to the handling of non-normality in the data patterns. Many financial series fall into that category and data distributions produce behaviours that fall outside normal laws. Mixing standard time series models with heavy tail processes brings the remedy and produces new class of hybrid series that capture the non-normal patterns reasonably well.
I have attached the notebook where I copied the code from the original application.
Hope this helps.
Thanks Igor! This helps a lot!
may I know how to define the param and lastval parameters to generate the ARMA models with heavy tails?
Would it possible as well to know the code to create the graphs comparing the volatilities of the ARMA and the heavy tail models?
By the way, for the lansim and levysim process definitions, shouldn't we use lanvol and levyvol respectively instead of stTvol? Many thanks in advance.