predicted rate from a random intercept poison model

[BayzidurRahman]

I am fitting the following model with my count data.

proc glimmix data=tre;
class ward mrn period sex;
model tre_count=period period*sex sexaknt1-aknt3 / dist=poisson link=log offset=logord;
random intercept / subject=ward;
run; 

How can I add a new variable in the existing dataset with predicted rate from the fitted model?

[Ksharp]

data insure;
input id n c car$ age;
carnum=1;
if car="medium" then carnum=2;
if car="large" then carnum=3;
ln = log(n);
datalines;
1 500 42 small 1
1 1200 37 medium 1
1 100 1 large 1
1 400 101 small 2
1 500 73 medium 2
1 300 14 large 2
2 50 4 small 1
2 10 7 medium 1
2 300 10 large 1
2 420 101 small 2
2 510 73 medium 2
2 30 14 large 2
3 70 40 small 1
3 150 17 medium 1
3 30 20 large 1
3 250 10 small 2
3 50 3 medium 2
3 300 140 large 2
;


proc glimmix data=insure;
class car age;
model c=car age / dist=poisson link=log offset=ln;
random intercept / subject=id;
store gmxplants;
run; 
proc plm restore=gmxplants;
score data=insure out=out pred stderr lclm uclm / nooffset ilink;
run;

[BayzidR]

Thanks very much.
How can I report the regression coefficients with their 95% for the variables age and car?

[SAS_Rob]

The SOLUTION and CL options on the MODEL statement requests the parameter estimates and that t-type confidence limits be constructed for each of the fixed-effects parameter estimates. The confidence level is 0.95 by default; this can be changed with the ALPHA= option.

model tre_count=period period*sex sexaknt1-aknt3 / dist=poisson link=log offset=logord cl solution;

[BayzidR]

Thanks. Is there any option to obtain the IRR or exponentiation coefficients?

[StatDave]

All these questions, including the rate ratio, are addressed in this note. It is illustrated using PROC GENMOD, but the same basic methods apply.

[ballardw]

Short answer: You can't.

Reason, once a data set is created in SAS adding a variable requires REPLACING the data set. You don't "add" a variable to it. If you want the same name you replace the data set.

Basic approach would be to create a data set with the predicted rate. That would be the OUTPUT statement

Output out=newdatasetname  pred=nameofpredictedvaluevariable ;

Suggestion: DO NOT reuse the source data set name and replace it. If you make a mistake you may corrupt your existing data.

---

@BayzidurRahman wrote:

I am fitting the following model with my count data.

proc glimmix data=tre;
class ward mrn period sex;
model tre_count=period period*sex sexaknt1-aknt3 / dist=poisson link=log offset=logord;
random intercept / subject=ward;
run; 

How can I add a new variable in the existing dataset with predicted rate from the fitted model?

---