Good afternoon to everyone. I am working on a model with three sequential dependent variables, such that the occurrence of the second depends on the first, and the occurrence of the third depends on the second. I had considered using the seqlogit command to model this relationship.
The problem is that there is strong suspicion of reciprocal causality between the sequential variable and my main independent variable. I had thought about addressing this using an instrumental variables (IV) methodology, but I have read that this command does not support it.
Therefore, I had considered estimating the probability of my independent variable using a logit model, and then using that prediction as an independent variable in the sequential model:
Is this econometrically correct?
The problem is that there is strong suspicion of reciprocal causality between the sequential variable and my main independent variable. I had thought about addressing this using an instrumental variables (IV) methodology, but I have read that this command does not support it.
Therefore, I had considered estimating the probability of my independent variable using a logit model, and then using that prediction as an independent variable in the sequential model:
HTML Code:
logit x1 x2 xn z predict x1_hat, p seqlogit y x1_hat x2 xn, tree(...)
