Postestimation tests with xtsur

[Fabian Villarreal]

Dear Statalist,

I am trying to estimate a regression with Panel data SUR method with the command xtsur, since I have more time periods than individuals. In summary, I have two equations:

xtsur (eq1: y1 x1 x2 x3 ...) (eq2: y2 x1 x2 x3 ...). I have obtained good statistical significance in the coefficients, but I would like to perform some robustness tests.

I would like to perform some postestimation tests with respect to the normality of the residuals, autocorrelation, homocedasticity, and so on. Hence, I tried to use the command: predict to estimate the residuals (predict ehat, residuals) (I also tried the following predict ehat1, residuals equation(eq1)... and the same for equation 2). Nonetheless, I have gotten the following message: "option residuals not allowed". Does anyone know how to do postestimation tests with xtsur?

Or more generally, how to do postestimation tests with the residuals after multiple-equation (ME) estimators?

Thanks.

[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.

With user-written estimators, you often don't get all the features Stata includes. They are not guaranteed to play well with other parts of Stata. If you can just predict the dv, you can calculate the residuals.