0

I am performing deviance goodness-of-fit test on my model (used negative binomial regression), and the R summary() table of my model gives the following:

Call:
glm.nb(formula = topPagesCount ~ DB_LEGAL_STATUS_CODE_V2 + BCORP_INDUSTRY_DENSITY + 
DENSE_MSA + DENSE_MSA * BCORP_INDUSTRY_DENSITY, data = greatDF, 
control = glm.control(maxit = 100), init.theta = 0.2103980872, 
link = log)

Deviance Residuals: 
Min       1Q   Median       3Q      Max  
-1.9154  -1.2815  -0.7786  -0.2994   2.3033  

Coefficients:
                                    Estimate Std. Error z value Pr(>|z|)    
  ( Intercept)                      4.4117492  0.4566746   9.661  < 2e-16 ***
  DB_LEGAL_STATUS_CODE_V212        -0.6420125  0.3006310  -2.136  0.03272 *  
  BCORP_INDUSTRY_DENSITY            0.0067614  0.0092156   0.734  0.46314    
  DENSE_MSA                         0.0311327  0.0104588   2.977  0.00291 ** 
  BCORP_INDUSTRY_DENSITY:DENSE_MSA -0.0006298  0.0002075  -3.035  0.00241 ** 
   ---
  Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

  (Dispersion parameter for Negative Binomial(0.2104) family taken to be 1)

 Null deviance: 393.12  on 304  degrees of freedom
 Residual deviance: 362.31  on 300  degrees of freedom
 AIC: 2794.2

 Number of Fisher Scoring iterations: 1


          Theta:  0.2104 
      Std. Err.:  0.0158 

  2 x log-likelihood:  -2782.1530

the dispersion parameter of 0.2104 is getting me little confused...is it supposed to be a good thing? or is this a bad thing?

thank you

Jin-Dominique
  • 831
  • 2
  • 9
  • 18
  • @Peter: As this is closed as a duplicate, would you mind moving my answer? Gavin's answer explains what the dispersion is; mine how the negative binomial distribution is parameterized: so they complement each other. – Scortchi - Reinstate Monica Nov 25 '13 at 16:04

0 Answers0