Hello,

I would be grateful if someone in the list can advise in the following problem, I post it for the second time, may be in the first time I did not explain my problem very well

The problem that I have is the following, I have a fertility data with region (150 regions), group ( women age1 20-30 and women age2 31-40 ) , number of children by region, number of females by region.

I would like to model the fertility ratio (child/female) as a function of the group and region

using %glimmix:

In the first step I consider a random intercept model

l have noticed that the variance of the random effect is "Region 0.02168", but I do not have the significance test for this parameter.

I would like to ask the following questions:

1- If my binomial distribution is appropriate in my case?

2- How I can obtain the significance test, for the variance ?

3- How I can obtain the residuals and their corresponding confidence interval?

4- Can use instead of %glimmix , the NLMIXEd procedure, since I have some experience using the proc nlmixed?

Thank you in advance for any help

Adel

My code:

%glimmix(data=fertil.mydatacl1,

stmts=%str(class region group ;

model child/female=group/solution cl;

random region ;

parms (0)(1.0) /EQCONS=2;),

error = binomial,

link=logit,

procopt=order=data);

run;

--------------------------------- Appel audio GRATUIT partout dans le monde avec le nouveau Yahoo! Messenger Téléchargez le ici !