multivariate regression

[Maria Sassi]

I am running a multivariate regression. Unfortunately, the Breusch-Pasan test indicates residuals of the dependent variables not independent of each other. I would be very pleased if some can send me the codes for performing a robust estimate.
Thank you and regards.
Maria

[Carlo Lazzaro]

Maria:
I'm under the impression that you ran a multiple -regress-, not a -mvreg- one.
That said, if you detected serial correlation in the residual distribution, the easiest fix is to invoke -cluster- standard errors.

[Joro Kolev]

Maria, "the Breusch–Pagan test [and note the spelling of the second author's name], developed in 1979 by Trevor Breusch and Adrian Pagan, is used to test for heteroskedasticity in a linear regression model. "

I am guessing as Carlo that you did not run multivariate regression -mvreg-, but rather multiple regression -regress-. Just add the option robust.

reg y x, robust.

[Maria Sassi]

Thank you for your answer. What I did was to write the command
mvreg, corr
as indicated in https://www.stata.com/manuals13/mvmvreg.pdf
Now, I am pretty confused about the meaning of this test.
Regards.
Maria

[Eric de Souza]

Maria Sassi: the question is whether you wanted to estimate the coefficients of one equation with one dependent variable (say y) and several explanatory variables (say x1, x2, x3, ...); or whether you wanted to estimate the coefficients of several equations, in which case you have several dependent variables.
Please write down the full command you issued with the variables

[Maria Sassi]

Dear Eric,
Thank you so much for your support.
Here the full commands I used
manova Protein Vitamin Hem_iron = c.DDS c.health_env literacy c.HHae resi Children_sick Chronic_prob c.workers_hhmembers
mvreg, corr

[Joro Kolev]

This is multivariate regression indeed. And this Breusch–Pagan test is a test of the null that the residuals of the dependent variables are independent
of each other. In other words, if your p-value is small, you do have a multivariate regression, and you cannot analyse the equations separately.

You do not need to do anything in particular, just carry on with your multivariate analysis.

[Maria Sassi]

Thank you so much indeed.