Anat (Manes) Tchetchik <anatmanes@gmail.com>:
I said nothing of the kind, but it is true that the conditional mean
of the outcome y must always be positive.
But outcomes can be zero or even negative.
The assumption is that
ln(E(y|X) is linear in X
not that y is never zero or even negative.
Of course ln(.) must have a positive argument.
On Wed, Mar 6, 2013 at 4:47 PM, Anat (Manes) Tchetchik
<anatmanes@gmail.com> wrote:
> Thanks Austin, you mean that the conditional mean must always be positive?
>
>
> On Wed, Mar 6, 2013 at 11:31 PM, Austin Nichols <austinnichols@gmail.com> wrote:
>> Anat (Manes) Tchetchik <anatmanes@gmail.com>:
>> No assumption about conditional variance is required for consistent
>> estimation of coefficients in a Poisson model or -glm- with a log
>> link. Only the conditional mean assumption is required.
>>
>> The FAQ cited contains many errors--if anyone knows who wrote it,
>> please them know.
>>
>> On Wed, Mar 6, 2013 at 4:01 PM, Anat (Manes) Tchetchik
>> <anatmanes@gmail.com> wrote:
>>> Dear all,
>>> I'm still struggling with my IV Poisson model.
>>> I'm not sure; does the assumption about conditional variance for
>>> consistency is required ?
>>> I read in http://www.ats.ucla.edu/stat/stata/dae/poissonreg.htm that
>>> the Poisson has a strong assumption; that the conditional variance
>>> equals conditional mean
>>> If so then what test should be used to verify if my data is
>>> "qualified" for using this model?
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/faqs/resources/statalist-faq/
* http://www.ats.ucla.edu/stat/stata/
