I run a three-level nested model as following,
- repeated measures of individual, and individual nested within couples
- IV and DV are both binary variables, moderator Pcsc is continuous variable (centered)
And I would like to call for the simple slope and intercept estimates like what people can do in PROC MIXED Estimate statement, by assigning values of M-SD and M+SD to the moderator, following,
proc nlmixed qpoints=20 tech=newrap;
eta = a0+a1*Propartner_lagged+a2*Pcsc++a3*Pcsc*Propartner_lagged+PE+CE;
mu = exp(eta) / (1+exp(eta));
model propartner_dummy ~ binary (mu);
random CE ~ Normal (0, sc) subject = couple_Id;
random PE ~ Normal (0, sp) subject = id (couple_Id);
estimate 'PreviousBeh effect at low Pcs' Pcsc=7.52 Pcsc*Propartner_lagged=7.52;
estimate 'PreviousBeh effect at high Pcs' Pcsc=9.74 Pcsc*Propartner_lagged=9.74;
run;However, the programme report error as followings,
Is it because of the logit link transformation in proc nlmixed? Then how to do the simple slope analysis to interprect the interaction? I'm quite new at handing multilevel modeling, please help me 🙂
PS I'm using SAS/STAT 13.2 because I need multiple RANDOM statements to model the three levels.
Thanks a lot!
