Hi,
I am trying to estimate a system of 5 probit models being the dependent binary variables w1 w2 w3 w4 and y.
I would like to estimate the system such that the correlation of the error terms between each couple of (wj,wi) is zero, but the correlation (wj,y) is unrestrained. Basically, my system is supposed to produce 4 more parameters (correlations) than 5 independent probit models.
I am working with gsem and mvprobit, but I am struggling to impose these restrictions.
Any help would be greatly appreciated,
Simone
I am trying to estimate a system of 5 probit models being the dependent binary variables w1 w2 w3 w4 and y.
I would like to estimate the system such that the correlation of the error terms between each couple of (wj,wi) is zero, but the correlation (wj,y) is unrestrained. Basically, my system is supposed to produce 4 more parameters (correlations) than 5 independent probit models.
I am working with gsem and mvprobit, but I am struggling to impose these restrictions.
Any help would be greatly appreciated,
Simone
