Re: [Gretl-users] HOW TO CORRECT FOR HETEROSCEDASTICITY WHEN DEALING WITH NON POSITIVE NUMBERS IN EVIEWS
On Fri, 9 Oct 2015, Allin Cottrell wrote:
On Fri, 9 Oct 2015, Sven Schreiber wrote:
> Am 07.10.2015 um 16:45 schrieb Nathan Nyamapasi:
>> Good day
>>
>> may you kindly assist by telling me how i can correct for
>> heteroscedasticity when dealing with data that is composed of non
>> positive values like zero and negative value. your urgent assistance is
>> greatly appreciated in advance
>
> Nathan, that sounds very much like a question that is not gretl-related,
> even when I abstract from the strange word "Eviews" in your subject.
One further thought: if the background to the question was the
common textbook suggestion, "In case of heteroskedasticity, try
re-specifying the model in log form", then the answer (not
gretl-specific in any way) is simply, you cannot do that if the
variable(s) of interest have non-positive values. Although... I've
seen cases where an econometrician will add a positive constant to a
series so that its log can be taken...
Allin Cottrell