Nested random effect in NLMIXED

[boban]

Hi,

I’m using proc nlmixed to analyse a data set with nested random effects. Ontsurv is the binary response variable, and is nested under both mom and dad (class variables). I used the following program but it doesn’t seem to work.

proc nlmixed data = final Qpoints = 2;

array aeffect { 1 };

array beffect { 1 };

vara = exp(2*logsiga);

varb_a = exp (2*logsigb_a);

mean = intercept + aeffect {1} + beffect {dad};

parms B0 = 10 V_u0 = 80;

logodds = B0 +u0j;

odds = exp (logodds);

P = odds/(1+odds);

model ontsurv ~ binary(p);

random aeffect1 beffect1 ~

normal ( [0,0],

         [vara,

            0, varb_a] ) subject = mom;

estimate 'var[mom]     ' vara;

estimate 'var[dad (mom)] ' varb_a;

run;

Does anyone know how to fix the problem, either by changing the syntax or with a good reference that I could look into?

Cheers,

B