Re: [Gretl-users] Test reverse causality
Il giorno gio, 10/04/2014 alle 11.04 +0200, andrea_conti_86(a)tiscali.it
ha scritto:
Dear Gretl users,
i'm doing panel regression where i suppose there's a reverse causality
among independent and dependent variables and I can't find instrument
for them .
Does exist a test that tell me if such reverse causality exists?
Dear Andrea,
assume that two barmen work in a bar, in different nights, and they both
prepare the same cocktail made of gin and lemonade. But one fills more
or less half of the glass with gin, fills it with lemonade until it is
exactly full, and then stirs. The other fills more or less half of the
glass with lemonade, fills it with gin until it is exactly full, and
then stirs.
The cocktails produced by the two barmen have, on average, exactly the
same amount of alcool, lemon and any other chemical component.
By only analyzing the cocktails produced during a given night, you can
presumably identify a negative relationship between the amount of alcool
and the amount of lemon. But intuitively, do you think that Gretl, or
any statistical software, could ever tell you which barman is working
tonight (that is, in which sense the causality is going)?
Pietro