Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
st: comparing more than 2 negative binomial models using fitstat command
From
"Cale, Grace E" <[email protected]>
To
"[email protected]" <[email protected]>
Subject
st: comparing more than 2 negative binomial models using fitstat command
Date
Tue, 25 Mar 2014 20:08:10 +0000
Good afternoon!
I am attempting to compare AIC and BIC' scores for four negative binomial regression models I am using for my thesis.
I am using Stata/IC for Windows 7 (32-bit).
For the sake of your time, my simple question is: what is the best way to compare more than two nested negative binomial models' AIC and BIC' scores?
For details:
I have been using the -nbreg- command for a set of nested models, and have successfully compared AIC and BIC' scores between the two models with the below commands:
I type: -quietly nbreg polact model2independentvariables if cooksd!=1, vce(robust)-
And: -quietly fitstat, saving(m2)-
Then: -quietly nbreg polact model1independentvariables if cooksd!=1 & e(sample)-
And: - fitstat, using(m2)-
This gives me:
.Measures of Fit for nbreg of polact
Current Saved Difference
.Model: nbreg nbreg
.N: 732 732 0
.Log-Lik Intercept Only: -1531.264 -1531.264 0.000
.Log-Lik Full Model: -1424.746 -1413.334 -11.412
.D: 2849.492(707) 2826.668(704) 22.824(3)
.LR: 213.037(23) 235.861(26) -22.824(-3)
.Prob > LR: 0.000 0.000 0.000
.McFadden's R2: 0.070 0.077 -0.007
.McFadden's Adj R2: 0.053 0.059 -0.005
.Maximum Likelihood R2: 0.253 0.275 -0.023
.Cragg & Uhler's R2: 0.253 0.275 -0.023
.AIC: 3.961 3.938 0.023
.AIC*n: 2899.492 2882.668 16.824
.BIC: -1813.725 -1816.762 3.037
.BIC': -61.334 -64.371 3.037
.Difference of 3.037 in BIC' provides positive support for saved model.
But when I compare models two and three, both sets of AIC and BIC' scores are different, so I cannot include these scores in my analysis tables, and the -fitstat- command does not allow for more than two tables to be compared. Do you have any recommendations? I will gladly provide additional details through this discussion list, or privately via email: [email protected]
Thank you,
Grace Cale
Teaching Assistant
MA Program: Sociology
University of Kentucky
Email: [email protected]
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/faqs/resources/statalist-faq/
* http://www.ats.ucla.edu/stat/stata/