Hi All
I'm running margins post multinomial logistic regression (imputed data), and as recommended I'm using the 'mimrgns' package:
The model:
The code for margins (I'm primarily interested in interaction effects here):
I have a couple of questions:
- I'm interested in obtaining 'predicted probabilities' for the outcome 'suicide' based on interacting categories of 'comorbid' and 'sexualx'. Have I specified the right options for this, i.e.: predict(pr) and vce(unconditional)? What exactly does the latter do? And as with regular 'margins; run after a regression model, these margins can be interpreted as the proportions of individuals with the outcome?
-The 95% CIs estimated for the margins: Are these CIs interpreted the same way as in the model above? i.e., it helps us understand significant differences between the predicted margin from the reference category? I assume not, but we can compare all margins against one another using their CIs?
Many thanks!
/Amal
I'm running margins post multinomial logistic regression (imputed data), and as recommended I'm using the 'mimrgns' package:
The model:
Code:
mi est, post dots or: svy: logit suicide i.comorbid##i.sexualx i.sex i.ethnic2cat ib5.incomeq3
Code:
mimrgns (comorbid##i.sexualx), predict(pr) vce(unconditional)
- I'm interested in obtaining 'predicted probabilities' for the outcome 'suicide' based on interacting categories of 'comorbid' and 'sexualx'. Have I specified the right options for this, i.e.: predict(pr) and vce(unconditional)? What exactly does the latter do? And as with regular 'margins; run after a regression model, these margins can be interpreted as the proportions of individuals with the outcome?
-The 95% CIs estimated for the margins: Are these CIs interpreted the same way as in the model above? i.e., it helps us understand significant differences between the predicted margin from the reference category? I assume not, but we can compare all margins against one another using their CIs?
Many thanks!
/Amal
