SUREG with analytic weights

[Miguel Henry]

Hello,

I am considering to use sureg with analytic weights where the set of regressors differ in each equation. However, sureg in Stata allows only for a single set of analytic weights to be used across all equations (all dependent variables). The problem is that I have different sets of analytic weights for the different equations, and sureg does not allow for that. Any suggestions will be appreciated, if any.

[George Ford]

Could force it by multiplying every variable and the constant by sqrt(wgt) [which means a no constant regression with sqrt(wgt) as a regressor].

Might also try suest instead, but sometimes it gets upset with weights too.

[Miguel Henry]

Thank you George. Transforming all the variables in each equation by diving them by the weights, including the inverse weight as variable, and using the nocons option to get rid of the constant term seems to be the only way to do it these days. It would be great if in the near future StataCorp could allow for more than one set of weights to be used across equations when using sureg.