|
|
 |
| SEM with binary indicators |
 
|
| Message/Author |
|
|
|
|
The model has 5 latent variables, and each has multiple binary indicators.
Q1) I wonder if I need to treat all binary indicators as categorical or not.
Q2) The model below does not converge. I wonder what I should do to make it work. When binary indicators were treated as continuous (excluded from "categorical are"), then it converges. But I am not sure if the numbers I get are correct or not in this case.
Please advise.
[syntax]
Variable:
...
categorical are
intent know1-know10 risk1-risk9 faci1-faci11 bar1-bar15 cn1-cn3;
...
Analysis:
type=general;
estimator=WLSMV;
iterations=1000;
parameterization=theta;
bootstrap=5000;
Model:
CN BY
cn1 cn2 cn3;
KNOW BY
know1-know10;
PR BY
risk1-risk9;
PF BY
faci1-faci11;
PB BY
bar1-bar15;
intent ON
PR (pr_int)
PF (pf_int)
PB (pb_int)
KNOW (know_int)
CN (cn_int);
PR ON
CN (cn_pr)
KNOW (know_pr);
PF ON
CN (cn_pf)
KNOW (know_pf);
PB ON
CN (cn_pb)
KNOW (know_pb);
PR WITH PF PB;
PF WITH PB;
CN WITH KNOW; |
|
|
|
|
Q1: I would recommend that.
Q2: Check if it works when you are not requesting bootstrapping.
If this doesn't help, send your full output and data to Support along with your license number. |
|
|
|
|
Thank you. I emailed but the support contact for my M+ expired.
I modified the syntax based on your advice. Still, the model does not converge. Here are some details of what I have done so far.
- All binary indicators are considered as categorical as advised
- Listwise deletion is on
- I tried it without bootstrapping as advised but still no convergence
- Without listwise deletion, it converges. However, with listwise deletion=on (FIML) & bootstrapping=100, I get 999 for all SE.
- I tried bootstrapping=1000, but it gets stuck around 130 for some reason I am not sure.
- I am using bootstrapping to get bias-corrected CI for indirect effects.
I would really appreciate your advice on this problem. |
|
|
|
|
I would not use the listwise deletion option. First make sure it converges without bootstrapping.
To give you more advice we need to see the output and data and for that support you need to have a support contract. |
|
|
| Back to top |
|
|
