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 at the end of May, and its replacement, statalist.org is already up and running.


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

Re: st: Regression Equation in ZINB regression


From   Partho Sarkar <partho.ss+lists@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Regression Equation in ZINB regression
Date   Tue, 20 Sep 2011 22:02:14 +0530

Rachel

You are very welcome- and even I got to learn something new through
this!  Incidentally, I notice now some typos in my messages, but I
think from the references I gave you will get a clear enough idea of
how to interprete(emphasis) your results- and I would imagine that is
what would be expected from a dissertation in biology.

Best
Partho

On Tue, Sep 20, 2011 at 9:30 PM, rachel grant <rachelannegrant@gmail.com> wrote:
> Thanks Partho for taking the time to help me.
> It is becoming clearer. I am going to look into it further and I'll
> come back to the list if I need more help
> thanks again
> Rachel
>
> On 20 September 2011 12:50, Partho Sarkar <partho.ss+lists@gmail.com> wrote:
>> Oops, sorry again!  My apologies- I am doing this in between other
>> work.  I meant to write:
>>
>> E[yi|xi]=count for x=xi =lambda[i]=a0+a1xi + etc, i.e., the
>> probability lambda is different for different levels of xi.
>>
>> and
>>
>> L(y1,y2,..,yn)=F(lambda[1],lambda[2],...,lambda[n])=G(x1,x2,...,xn)
>>
>> Partho
>>
>>
>> On Tue, Sep 20, 2011 at 5:09 PM, Partho Sarkar
>> <partho.ss+lists@gmail.com> wrote:
>>> Rachel, it seems you posted the same question under 2 headers- I just
>>> replied on the thread: Regression Equation for Zero inflated negative
>>> binomial.  I wanted to make a slight correction-  I said the parameter
>>> lambda is is the LHS in the regression, which is not quite correct.
>>> The mean (or expected) count is equal to lambda.  So we really use the
>>> fact that :
>>>
>>> E[yi|xi]=count for x=xi =lambda=a0+a1xi + etc.
>>>
>>> to get the "likelihood" of observing all the yi's as a function of the
>>> observed xi's:
>>>
>>> L(y1,y2,..,yn)=F(lambda)=G(x1,x2,...,xn)
>>>
>>> and then maximize this likelihood function to estimate lambda.
>>>
>>> Hope this helps
>>>
>>> Partho
>>>
>>> On Tue, Sep 20, 2011 at 4:19 PM, Partho Sarkar
>>> <partho.ss+lists@gmail.com> wrote:
>>>> My post crossed yours in the mail!  Hope you can manage now.
>>>>
>>>> Partho
>>>>
>>>> On Tue, Sep 20, 2011 at 4:03 PM, rachel grant <rachelannegrant@gmail.com> wrote:
>>>>> Thank you Partho. Thank you all for your patience.
>>>>> I have now understood that I cannot get Stata to do this and I need to
>>>>> reconstruct by inserting my co-efficients into the equation. Fine
>>>>>
>>>>> Now the problem I have is finding a general equation for ZINB in order
>>>>> to reconstruct.
>>>>> Is the general equation for ZINB the same as Poisson? I cannot find
>>>>> anywhere a general equation for ZINB.
>>>>> I could use a general log-linear Poisson type equation but would this
>>>>> be correct? and how do I deal with the inflated part, is this merely a
>>>>> seperate log-linear eqn, or is it incorporated into th general ZINB
>>>>> eqn. Thanks, Rachel
>>>>>
>>>>> On 19 September 2011 08:13, Partho Sarkar <partho.ss+lists@gmail.com> wrote:
>>>>>> Hi Rachel
>>>>>>
>>>>>> Just to clarify: in order to actually get the "regression equation",
>>>>>> by which I assume you mean something like
>>>>>>
>>>>>> yhat=12.56+2.6*x1+0.34*x2 ,
>>>>>>
>>>>>> you do need to reconstruct the equation form the output displayed (as
>>>>>> exemplified by Anees's results).  You could see this page for an
>>>>>> example: http://dss.princeton.edu/online_help/analysis/interpreting_regression.htm
>>>>>> .
>>>>>>
>>>>>> With a little more programming effort, this process could be automated
>>>>>> using e-class macros etc.  (There is very probably a user-written
>>>>>> program to do this somewhere out there- -estadd- from here
>>>>>> http://repec.org/bocode/e/estout/ might help partly, and you can
>>>>>> search the rich REPEC archives).
>>>>>>
>>>>>> (In this sense Stata may not be as "user friendly" as Minitab or even
>>>>>> some more ambitious  other packages, which helpfully report the exact
>>>>>> "prediction equation" in the form you might want, but then it can do
>>>>>> so much more besides!)
>>>>>>
>>>>>> Hope this helps
>>>>>>
>>>>>> Partho
>>>>>>
>>>>>>> On Mon, Sep 19, 2011 at 12:34 AM, rachel grant <rachelannegrant@gmail.com> wrote:
>>>>>>>>
>>>>>>>> I have searched the help files but I cannot understand how to get
>>>>>>>> Stata to display the regression equation.
>>>>>>>> If anyone is able to help, without referring me to manual, it would be
>>>>>>>> really appreciated.
>>>>>>>>
>>>>>>>> --
>>>>>>>> regards, Rachel Grant
>>>>>>>> *
>>>>>>>> *   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/
>>>>>>
>>>>>
>>>>>
>>>>>
>>>>> --
>>>>> regards, Rachel
>>>>>
>>>>> *
>>>>> *   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/
>>
>
>
>
> --
> regards, Rachel
>
> *
> *   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