st: Constrained Regression with lagged variables

[Alfonso Sánchez-Peñalver <alfonso.statalist@gmail.com>]

Hi StataList,

I want to use the OLS transformation strategy explained in Sean Becketti’s Introduction to Time Series Using Stata (Stata Press), to estimate a model with AR(1) consistent standard errors. In the book Sean performs the regression without constraining the coefficients, but it follows from the explanation it follows that we should.

To illustrate the model is

yt = (1 - rho)b0 + rho yt-1 + b1 xt - rho b1 xt-1 + et.

If we specify the model then as 

yt = a0 + a1 yt-1 + a2 xt + a3 xt-1 + et

it follows that a3 = - a1 * a2, so we should be constraining it. Looking through the FAQ I found that -cnsreg- should be used for linear constrained regression. So i defined my constraint as

constraint 1 L.x = - L.y * x

and run the following command:

cnsreg y L.y x L.x, constraints(1).

I get the following error:

note: constraint number 1 caused error r(131).

I am wondering if this is because of the use of lagged prefixes, or because of the constraint involving the multiplication of two coefficients. Can someone through some light into this and also how I can go about constraining the coefficients accordingly.

On a side note, why doesn’t regress accept constraints?

Thanks,

Alfonso
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/