Extended Regression model with endogeneous covariate in main equation and selection equation

[naveed ahmed]

Dear Statalisters,

Using stata /IC 15.1, I am trying to run the following command:

eregress y x1 x2 x3, endog(x2=z1 x3) select(selected= x2 x3 w1)

When I run this command STATA reports that "initial values not feasible", however, when I add nomain with endog(x2=z1 x3,nomain) and don't include x2 in the main equation stata reports results .However, I really need to include x2 in the main equation. Any advice would be appreciated.

regards,
Naveed

[Phil Bromiley]

You'll increase your chances of a helpful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex. Really diagnosing such data-specific problems is easier with data.

You can try out other starting values or other maximization commands. You might try cmp or GSEM as well. What happens if you drop x2 from the selection equation?

[Obaid Ur Rehman]

Check the following probabilities:

1. Round-off your "y" variable

2. if your endogenous covariates are categorical, then put i, b or c symbol with x1, x2, x3, z1 or w1 variable in the equation.

Hope this'd be helpful