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
- /
- Mianalyze GLM Analysis

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
- Highlight
- Email to a Friend
- Report Inappropriate Content

02-14-2015 11:02 AM

Hi,

I would like to know whether there is a syntax/command for combining GLM effect sizes using MIANALYZE. If it is, could you let me know where I could get it? Thank you so much.

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

02-16-2015 11:01 AM

Mmm. Haven't seen any macros (which is what it would take) on this one. The key here is that the effect size estimates will be different for each imputation, and I don't believe that eta squared or omega squared are "additive" in the sense that you could take the average of the imputational values as applying to the results from MIANALYZE. If I were searching for this on Uncle Google's website, I might use key terms like "model averaging effect size SAS" to see what might pop up.

Steve Denham

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

02-16-2015 11:13 AM

Combining Type-III Analyses from Multiple Imputations:

http://support.sas.com/resources/papers/proceedings14/1543-2014.pdf

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

02-16-2015 11:25 AM

That macro is quite good for getting the MIANALYZE results, but as far as effect sizes, there is still something missing. However, since it does calculate the mean squares, one could get the LSMEANS, and calculate effect sizes in a data step. Cohen's d would be simple, eta squared and omega squared somewhat more difficult.

And, as always, I have a difficult time moving from MIXED to MIANALYZE with things like repeated measures (getting the matrix to be of the correct form). I generally appeal to the fact that maximum likelihood estimates under missing at random are optimal, and that imputation really adds little in that situation.

Steve Denham