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: not concave poisson model


From   Maarten buis <maartenbuis@yahoo.co.uk>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: not concave poisson model
Date   Thu, 15 Jul 2010 09:17:56 +0000 (GMT)

--- On Thu, 15/7/10, Ali Khashan wrote:
> I am fitting a Poisson model with aggregated person-years
> two time dependent variables; age, calendar year. I also
> have three other variables in the model.
> 
> When I run the model, the first iterations are 
> have  the message (not concave) but they converge at
> the end. If I remove age from the model it converges with no
> (not concave) message at all. The age variable is in 2-year
> categories (14-15, 16-17,......until 46-47). Considering
> that the model converges at the end, do I have to worry
> about the results? If yes what should I look for?

That is not a problem that the initial itterations are not
concave as long as the final itteration is concave. You 
need to look at your data to see wether age and calendar
year realy capture something different, there are dataset
types where the two are exactly the same, if you get one
year older the calender year has increased by one year as 
well and vice versa. The estimates you get are then 
artificially identified by the fact that you grouped age 
in 2 year categories, which does not sound very confincing
to me. Additionally, even if the data is created so that
the effects of these variables are identified (i.e. your
dataset contains people from different birth cohorts) then
I would not add age as a linear variable. We know that a 
year increase from 14 to 15 is a very different thing than
a year increase from 44 to 45. I personaly like linear 
splines for including such non-linear effects, as it often
strikes the right balance between allowing for non-linearity
and giving interpretable coefficients, see -help mkspline-.

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/


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