Page 1 of 1

Fitting Distributions

Posted: Mon Sep 28, 2015 9:13 am
by proberts
Given a series of data, is there a simple way to fit a probability distribution in eviews? The descriptive statistics provide data on normality etc but is eviews able to indicate which distribution might best fit a series? In my case, I have a series with a large N that is close to normal but skewed and peaked. Thanks in advance.

Re: Fitting Distributions

Posted: Mon Sep 28, 2015 9:44 am
by startz

Re: Fitting Distributions

Posted: Mon Sep 28, 2015 9:51 am
by proberts
You can get part of what you want from distribution plots.
http://www.eviews.com/help/helpintro.ht ... 3ww242909
It appears that this command only creates the standard histogram for the series otherwise I have to specify the distribution. Does eviews provide any suggestions for the probability distribution that is the best fit?

Re: Fitting Distributions

Posted: Mon Sep 28, 2015 9:52 am
by startz
I don't think so.

Re: Fitting Distributions

Posted: Tue Sep 29, 2015 8:52 am
by startz
Come to think of it, there is one more thing you might do. You can use View/Descriptive Statistics and tests/Empirical Distribution Tests to see if one of the available distributions is a good fit.

Re: Fitting Distributions

Posted: Tue Sep 29, 2015 5:37 pm
by NicolasR
Or you can estimate the parameters of a distribution by ML and compare the information criteria with other estimations, in case that you do not find the distribution you want in the option that startz suggests.

Code: Select all

wfcreate u 1 1000 genr z=nrnd*2+5 logl normal normal.append @logl logl1 normal.append res=z-c(1) normal.append var=c(2) normal.append logl1=@log((1/(var*@sqrt(2*@acos(-1))))*@exp((-1/2)*(res/var)^2)) normal.append @temp res var logl1 normal.append @param c(1) 1 c(2) 1 c(3) 1 normal.ml