Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: ZIP with Sample Selection


From   James McJoseph <james.mcjoseph@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: ZIP with Sample Selection
Date   Wed, 1 Dec 2010 11:24:29 +0200

Thanks, for this Maarten. I will sleep over it.

On Wed, Dec 1, 2010 at 10:46 AM, Maarten buis <maartenbuis@yahoo.co.uk> wrote:
> --- On Wed, 1/12/10, James McJoseph wrote:
>> 2. Is it appropriate to apply a count model to a data where
>> the only count present in the data are: 0, 1, 2.  Counts above
>> 2 are possible but very extremely rare. Another consideration, the
>> occurance of counts are not equally probable. 0 is more probable
>> than 1. 1 is more probable than 2, Hence 3 and above are much less
>> probably to even unlikely.
>
> Sounds consistent with a Poisson distribution with a extrememly low
> mean. One problem I can imagine is that in your subject line you
> refer to a zero inflated poisson. A regular Poisson with such an
> extremely low mean will have a huge amount of zeros. I can imagine
> that Stata will have trouble distinguishing between a zero inflated
> Poisson distribution and such an extreme Poisson. Add a selection
> equation to that, and I doubt whether you'll ever get convergence.
> Think of it this way: what we do when we do a statistical analysis
> is to compare people (or whatever your unit of analysis may be). If
> everybody where the same, such a comparsion would be very boring.
> In other words it is the differences between people that contains
> the information we use in our analysis. If you have such an extreme
> Poisson distribution, then your variance is likely to be very low,
> which means there is very little information present in your data.
> A consequence of that is that even if the dataset is large, you can
> still only estimate relatively simple models and certainly not a ZIP
> with sample selection, as both the ZIP part and the sample selection
> part require lots of information on their own, let alone in
> combination.
>
> Hope this helps,
> Maarten
>
> --------------------------
> Maarten L. Buis
> Institut fuer Soziologie
> Universitaet Tuebingen
> Wilhelmstrasse 36
> 72074 Tuebingen
> Germany
>
> http://www.maartenbuis.nl
> --------------------------
>
>
>
>
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
>

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index