Forum: help

linearhypothesis testing with SUR and HAC vcov [ Reply ]
By: Julien Mc Donald-Guimond on 2021-01-22 15:05

Hi,

First, let me thank you for that great package, it works very well and is quite easy to use.

I'm estimating a system of 11 equations with SUR. Each equation has 21915 observations and 24 parameters.

I'm trying to run the following test:

linearHypothesis(sur, restrict, test = "Chisq", vcov. = kernHAC(sur, bw = 14, kernel = "Bartlett", prewhite = FALSE))

but I get the following warning:

In estfun.systemfit(x, ...) :
the columns of the returned estimating function do not all sum up to zero, which indicates that the wrong estimating function is returned

I've seen in the documentation that this is something that can happen, but would you say it renders the test invalid?

I'm not using any instruments, and the restrictions seem correctly specified since the test works fine when the vcov. argument is left to its default.

Thank you very much.
R version 4.0.2 (2020-06-22)
systemfit 1.1-24