Statalist


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

Re: st: RE: nbreg proc genmod equivalency


From   Steven Samuels <sjhsamuels@earthlink.net>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: RE: nbreg proc genmod equivalency
Date   Mon, 1 Dec 2008 07:50:18 -0500


On Nov 28, 2008, at 5:14 AM, Victor M. Enciso M. wrote:




I'm doing a negative binomial regression using Proc Genmod on SAS
where this is part of the output. I'm trying to reproduce this analysis in Stata but I don't know how to get the Type 1 and Type 3 Statistics. Anybody knows how to do this?


Type 1 Tests depend on the order of variables in the SAS Model statement. In your example the order was PAIR TREATMENT.

Type 1: Significance of
PAIR unadjusted
TREAT adjusted for PAIR

Type 3: Significance of
PAIR adjusted for TREAT
TREAT adjusted for PAIR

In Stata you can mimic the Type 1 Test for PAIR by fitting the Stata model with PAIR as the only predictor.. The Type 1 and Type 3 Test for TREAT is found by doing the Likelihood Ratio Test for the model with PAIR vs the model with PAIR and TREAT. See the -help- for - lrtest- , also for -nbreg postestimation-. Likewise, the Type 3 test for PAIR is the LRT test for the model with PAIR and TREAT vs the model with TREAT only.


Another question is why are the p-values of the contrast of tratment 1
wrt to treatment 2 (0.0296) and the one given for TRT (0.0336)
different?

The 0.0296 is the p-value of the Wald Chi Square Test (Est/SE)^2. The 0.036 is the p-value for the Likelihood Ratio Chi Square Test.


-Steve

On Nov 28, 2008, at 5:14 AM, Victor M. Enciso M. wrote:

I'm doing a negative binomial regression using Proc Genmod on SAS
where this is part of the output


Parameter   DF    Estimate       Error           Limits       Square
Pr > ChiSq

TRT   1      1     -0.2397      0.1102     -0.4557   -0.0237   4.73
0.0296
TRT   2      0      0.0000      0.0000      0.0000    0.0000
Dispersion   1      0.1268      0.0315      0.0651    0.1885

NOTE: The negative binomial dispersion parameter was estimated by
maximum likelihood.


                     LR Statistics For Type 1 Analysis
                          2*Log                 Chi-
          Source       Likelihood        DF     Square    Pr > ChiS

          Intercept     5189.2817
          PAIR          5253.0473        34      63.77        0.0015
          TRT           5257.5618         1       4.51        0.0336


                      LR Statistics For Type 3 Analysis

                                   Chi-
          Source           DF     Square    Pr > ChiSq

           PAIR             34      65.03        0.0011
           TRT               1       4.51        0.0336

*
*   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   |   What's new   |   Site index