Poisson regression with sample selection (heckpoisson)

[Anat Tchetchik]

Dear all,

I use Stata 15.1.
I have count data of 106 observations. The dependent var represents the number of days consumers of small hydroponics agricultural kit actually grew vegetables since buying it (the time passed since buying the kit is the exposure time). There is also a selection variable as some of the consumers abandoned the idea right after purchasing the kit and never used it.
Thus, poisson regression with sample selection -heckpoisson- deems appropriate.
I have 2 questions,
1. How can I specify upper limit (ul) as in the case of the tpoisson. There seems to be no option.
2. Currently all my efforts, including starting with only one or two variables, have failed. I receive constantly:

cannot compute an improvement -- discontinuous region encountered r(430);

. Any idea what might have cause this? is it the sample size? ( I have a variable to identify the selection equation).
Any help will be very much appreciated.
Best,
Anat

[Joao Santos Silva]

Dear Anat,

This looks more like a zero inflated model; one of the rare cases where such model would be meaningful.

Best wishes,

Joao

[Anat Tchetchik]

Thank you Joao,
I didn't thing about employing zero inflated Poisson model.. I guess it has to do with your comment .
I'll make some homework and give it a shot..
Thanks again,
Anat

[Anat Tchetchik]

Dear Joao,

I'm going over ZIP (zero inflation model) help document. I'm not sure what''s the meaning of the following quote:

inflate() equation that determines whether the count is zero

.
Do I need the specify the variables which may explain why the dependent var. is zero? (the word whether is a bit confusing..)
Is an "identifier" variable needed for the inflate equation? (a variable that appears in inflate( ) but not in the model?
Thanks again,
Anat

[Joao Santos Silva]

Dear Anat,

Those are the variables explaining whether or not the consumers abandoned the idea right after purchasing the kit and indeed it is good to have an "identifier", but that is not strictly needed.

Best wishes,

Joao

[Anat Tchetchik]

Thanks a lot Joao, highly appreciated!