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
- /
- SAS Programming
- /
- SAS Procedures
- /
- GEE GLIMMIX

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

02-21-2015 01:09 AM

Dear all,

i had done propensity score matching. i need use proc glimmix because permit to specify beta distribution for longitudinal model

How can account for matching variable in this proc ( the strata variable into conditional logistic)

*Beta GEE;

proc glimmix data=data_;

class id time trt;

model y = trt time time trt / dist=beta s ddfm=none covb;

were i account for strata (matching variable)??? :smileyshocked:

random _residual_ / subject=id type=cs vcorr;

lsmeans diab*time /ilink cl;

run;

Best

G Bonitta

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

Posted in reply to BonittaG

02-23-2015 02:01 PM

Some things to consider:

You need a name for the matching variable. For now, I will call it pairid.

There is a variable in the lsmeans statement called diab that is not identified in either the class or model statement. What does this variable represent?

Here is what might work:

proc glimmix data=data_;

class id pairid time trt;

model y = trt time trt*time/dist=beta s ddfm=none covb;

random intercept/subject=pairid;

random time/residual subject=id type= <> vcorr;

lsmeans trt*time/ilink cl;

run;

I did not specify the structure for the R side repeated in time variable. That will depend on the spacing on the 'time' variable and what assumptions can be made. Type=CS may involve too many assumptions to be a realistic covariance structure, although it is handy for fitting unequal spacing in time.

Steve Denham

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

Posted in reply to SteveDenham

02-24-2015 01:21 AM

Thank You very Much

you right pardon me my mistake diab is trt ....

I'm an R user and new into SAS enviroment..

Seeing your competence i ask agian a little question .. :

how can I plot mean residual vs decile of linear predictor about the specifided model we are discuss ?

Sorry for noise ...

Thank You again

Best

Bonitta

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

Posted in reply to BonittaG

02-24-2015 02:12 AM

For residuals plot that i need maby i can use this from

SAS/STAT(R) 9.2 User's Guide, Second Edition

output statement correct?

proc glimmix;

stuff code

output out=igausout pred=p resid=r;

run;

Best

Bonitta

Thank You

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

Posted in reply to BonittaG

02-24-2015 07:39 AM

Yes, the OUTPUT statement will give predicted and residual values. From there to your request for plotting is a journey through the DATA step to calculate the decile cutpoints, then the mean residual within those, and finally SGPLOT to plot the two against one another.

Steve Denham

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

Posted in reply to SteveDenham

02-25-2015 07:41 AM

Thank you ....can you so gentle to write the code for do that ? about the model

for plot mean residuals vs linear predictor

Thank You so much

Sorry for noise

Best

Bonitta

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

Posted in reply to BonittaG

02-26-2015 08:09 AM

I am not a coder in that way, so my attempts would be pretty horrible. Google for examples of each of the steps I mentioned, and I am sure you can find much better code than I can develop (or have time to develop).

Steve Denham