I would say that, if one is really interested in estimating the auxiliary model, she might estimate it without the need of using a function...
I agree with Artur: Allin's built-in commands are much better and versatile and there is no reason to keep the wooldridge_test_serial alive.
After having served gretl, it can now finally rest in peace :)
And, after all, I was very pleased to find that a little function I coded when I was studying Wooldridge's textbook and learning gretl's scripting language has been used by other people around the world.
Bye
Giuseppe
> @Giuseppe: I guess, this makes the wooldridge_test_serial package obsolete,
> right? Or does it cover some 'special' case?
The gfn function allows testing a panel-model _specification_ for
serial correlation, without actually estimating the model first. In
that respect it's like xtserial in Stata. Whether that's sufficient
reason to keep the package in play, I'm not sure.