topic Re: Proc surveylogistic regression goodness-of-fit test in Statistical Procedures

Proc surveylogistic regression goodness-of-fit test

shovakc — Fri, 11 Aug 2017 12:47:49 GMT

I am trying to find an appropriate goodness-of-fit test for proc surveylogistic procedure. Could you please help me with that? Thanks!

Re: Proc surveylogistic regression goodness-of-fit test

SAS_Rob — Fri, 11 Aug 2017 18:20:22 GMT

Currently, the only available goodness-of-fit tests in PROC SURVEYLOGISTIC are found in the default output in the Model Fit Statistics and "Testing Global Null Hypothesis: BETA=0" tables. This is really a limitation with logit models in general on complex survey data in that there are not a lot of measures that can be used to assess fit that are defined. The Hosmer-Lemeshow test, for example, has been shown to be inappropriate in complex sample designs. This is discussed in the following reference

http://www.amstat.org/sections/srms/proceedings/papers/1997_026.pdf

Re: Proc surveylogistic regression goodness-of-fit test

shovakc — Fri, 11 Aug 2017 18:29:03 GMT

Thank you for your reply. I tried clicking on the link for the reference and was redirected to this website: http://www.amstat.org/ASA/Membership/Sections-and-Interest-Groups.aspx

Is there another way to get this paper? Maybe it requires subscription to Amstat to access the paper. Thanks!

Re: Proc surveylogistic regression goodness-of-fit test

SAS_Rob — Mon, 14 Aug 2017 11:55:50 GMT

Looks liek the link has changed...try this one instead

http://ww2.amstat.org/sections/srms/Proceedings/papers/1997_026.pdf

Re: Proc surveylogistic regression goodness-of-fit test

shovakc — Mon, 14 Aug 2017 17:56:01 GMT

Thank you for the link. Do you know if SUDAAN or any other software that provides goodness-of-fit test for logistic regression of complex survey data?

Re: Proc surveylogistic regression goodness-of-fit test

shovakc — Tue, 19 Sep 2017 15:33:52 GMT

Are the statistics reported in the default output for model fitness adjusted to account for complex sample surveys? Would really appreciate that information.

Thanks!

Re: Proc surveylogistic regression goodness-of-fit test

SAS_Rob — Tue, 19 Sep 2017 16:11:53 GMT

No, those statistics I mentioned are all likelihood based which would not include any adjustments to the variance for the sampling.

http://support.sas.com/documentation/cdl/en/statug/68162/HTML/default/viewer.htm#statug_surveylogistic_details06.htm