Non-normal models

asked 2014-04-09 08:36:44 +0000

Pantera gravatar image

Hi all,

Background-problem: I have time series data which do not follow a normal distribution (according to skewnes and kartosis - or the Jarque-Bera test). The variables are equal to or greater than zero. The residuals are expected to be non-normally distributed.

I assume I need to use non-normal models. Shazam has a rich model-battery for non-normal models (see Chapter 21. Shazam manual). Question: Do the Shazam-expert-team or users of Shazam have any suggesitons of informative literature which could guide the user in the selection or search for the "right" model to model the non-normal distribution?

Regards,

edit retag flag offensive close merge delete

Comments

Usually this is done by looking at what distribution makes sense for the dependent variable based on the underlying theory (eg. economic) rather than the residuals.

SHAZAMHelp gravatar imageSHAZAMHelp ( 2014-04-11 06:22:38 +0000 )edit

Thanks for the comment, and I guess the literature mentioned in the chapter 21 could be a good starting point.

Pantera gravatar imagePantera ( 2014-04-11 17:08:40 +0000 )edit