03-28-2017 09:56 AM
I am trying to fi a weibull distribution to a list of frequency data, and I am using the following code in SAS 9.4
proc univariate data=data1;
histogram / weibull (Theta=est) midpoints=1 to 28 by 1;
I got my parameter estimates however SAS did not give me any Goodness of fit statistics. Is there a specific coding that allows me to get one?
03-28-2017 10:02 AM - edited 03-28-2017 10:03 AM
You code looks correct and should give GOF tables. Run the following and see if you see the "Goodness-of-Fit Tests for Weibull Distribution" table
ods trace on;
proc univariate data=sashelp.cars; var mpg_city; freq cylinders; histogram mpg_city / weibull (Theta=est) midpoints=10 to 60 by 2.5; run;
ods trace off;
The proposed code did not gice GOF tables unless c>2. I had to put in a value of theta to getW-sq and A-sq GOF statistics.
Is there another procedure available to fit a weibull distribution using point frequency data?
Yes, the documentation for the GOF tests in PROC UNIVARIATE mentions that the ECDF tests for GOF are only available when c > 2.
I'm not near a copy of Stephens and D'Agostino right now, so I can't look up why that constraint is required.