Goodness of fit for exponentiated Frechet

[Salah]

Hi

I want to know if SAS can do a goodness of test for other less famous distributions? Such as Frechet, exponentiated frechet, Kumaraswamy..etc.

Thank you

[Rick_SAS]

You posted the question the SAS/IML Community, so I assume you want to program these tests yourself? If so, the answer is yes. You can use the SAS/IML language to implement algorithms and statistical tests that are published in papers, textbooks, websites, etc.

It can be a challenge to know what GOF test to use. Some distributions have their own tests. Others use general criteria such as AIC or BIC.

In summary, you need to find the mathematical formula or algorithm that gives the GOF test, then implement it yourself. There are no built-in GOF tests in IML for obscure distributions.

[Salah]

Thank you so much for the advices.

Are there any built-in GOF tests in SAS in general?

Thank you

[Rick_SAS · Accepted solution]

Yes. Probably the SEVERITY procedure in SAS/ETS software provides functionality that is closest to what you want. It provides log-likeihood GOF measures (-2*LL, AIC, BIC,...) and ECDF measures (KS, AD, CvM,...).

You can use the FCMP procedure to define the PDF (or LOGPDF) of your distribution and then call it from PROC SEVERITY. There are several examples ni the doc that you can follow.