Dear Stata users,

I have the following issue: I have to estimate a system of simultaneous equations with non-linear constraints of form 0 < a < 1 using OLS.

Example 2 in https://www.stata.com/support/faqs/s...nstraints/#ex2 shows how to impose non-linear constraints of the form 0 < a < 1 using nl command that makes use of the inverse logit function.

In my case, I have to run a pair of regressions simultaneously with a non-linear constraint of the form 0 < [eqn1]beta + [eqn2]beta < 1.

Originally, I used reg3 and imposed a linear constraint of the form [eqn1]beta + [eqn2]beta = 1 but sometimes I got negative coefficients because I could not also impose the positiveness of the estimated coefficients.

So, I would like to know if there is a trick or a way you can suggest to run simultaneous regression equations with non-linear equations of form 0 < a < 1 without using any reparameterization of the coefficients as nl does?


Thank you,

Anna