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]

st: Hierarchial Poisson Model and Semidefinite Negative Hessian


From   Simone Peart Boyce <simonepeart@yahoo.com>
To   statalist@hsphsun2.harvard.edu
Subject   st: Hierarchial Poisson Model and Semidefinite Negative Hessian
Date   Wed, 1 Sep 2010 07:16:36 -0700 (PDT)

Hi,

I am trying to run a count data model on the effect of program participation 
on the number of suspensions received by a student.  I tried to run a 
hierarchial 

poisson model with students nested in schools, however, I received the below 
error message:

xtmepoisson oss jumpstart propcat2 propcat3 propcat4 propcat5 
    logit || HS_attend_dur: , var 

numerical derivatives are approximate
nearby values are missing
Iteration 0:   log likelihood = -1868.4358  (not concave)
Hessian is not negative semidefinite
r(430);

I thought this might be due to the high percentage of zeroes that I have in my 
dataset - over 75% -  that indicates that none of these students were suspended.
 
                            (sum) oss
-------------------------------------------------------------
      Percentiles      Smallest
 1%            0              0
 5%            0              0
10%            0              0       Obs                2449
25%            0              0       Sum of Wgt.        2449
50%            0                      Mean           .3891384
                        Largest       Std. Dev.      .9137992
75%            0              7
90%            1              7       Variance        .835029
95%            2              8       Skewness       3.341423
99%            4              8       Kurtosis       17.31081
 
However, I ran a similar regression using xtpoisson, normal re and I am 
able to obtain results.
 
xtset HS
xtpoisson oss jumpstart propcat2 propcat3 propcat4 propcat5 
    logit , re normal
 
What explains this discrepancy if the models should be equivalent?
 
The negative binomial might be better suited for my data given the high 
frequency of zeroes.  Is there any code for a hierarchial negative binomial?  
I wasn't able to find any.
 
Thanks in advance for your help,
Simone Boyce


      

*
*   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