[Gretl-users] Heteroskedasticity / correction

Sven Schreiber svetosch at gmx.net
Fri Jan 10 19:28:43 EST 2014


Am 11.01.2014 00:04, schrieb Ana Amaro ISG:
> When you multiply all the terms of the model by  1 / sqrt(x) the
> const of the original model turns out to be
> 
> const / sqrt(x)
> 
> So this term of the corrected model is now a parameter (const) times
> a variable (1/ sqrt(x)) meaning that a 'const' alone disapeared in
> the new model.
> 
> Is this clear?
> 
> 


I think I now understand what you mean; yes if x is a non-constant
variable --let me write this as x_i to make it clearer-- then of course
the inverse of its square root is not a constant, either. (Previously I
thought you were talking about the standard deviation of x, but you
weren't, sorry.)

However, you don't have to do this manually in gretl, you can use the
explicit weighted-least-squares functionality for that (menu: Model ->
Ohter linear models -> weighted least squares).

But as it happens, after doing WLS in gretl, it seems impossible (at
least in the GUI) to test again for remaining heteroskedasticity, via
another White test. Leaving philosophical discussions about repeated
testing aside, I wonder whether that should be enabled... (not a
question to Ana, but to Allin and Jack).

cheers,
sven


More information about the Gretl-users mailing list