turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

Find a Community

- Home
- /
- Analytics
- /
- Stat Procs
- /
- Generalized estimating equation output

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

03-19-2015 11:34 PM

Dear all,

How are you?

I am analyzing my data with only 2 continuous outcome repeated measures from each individuals(total=1320) with a GEE model with unstructured correlation structure.

I am not familiar with GEE but as far as I understand, QIC is for comparing the appropriateness of correlation structure and QICu is for variable selection, which the smallest the value the better.

Does it make it a significant predictor if the p-value of an effect <0.05 from the 'Score Statistics For Type 3 GEE Analysis' output? In my case, with or without corresponding p-value<0.05 effect doesn't impact the correlation too much. But including the corresponding effect increases the QIC and QICu though not too much. Hence I'm confused about whether to include or exclude such effect.

Your insight is greatly appreciated.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to wong

04-03-2015 01:38 PM

The score type3 test is a test of the predictor's association with the response. This is generally what you would use to assess the value of having the predictor in the model. It is not a test of its effect on the correlation among the measurements. But adding or removing predictors may effect the estimates of the correlation. The QIC statistics don't have a measure of variability, so it is not possible to say if adding or removing a predictor has a "significant" effect on QIC.