R-从glmmadmb输出中提取过分散参数

Question

I'm wondering if anyone knows of a way to extract the estimate for the dispersion parameter from the output of glmmadmb in R. I'm using a negative binomial model, and would like to use this code for several different species without having to go in and manually extract this value for the remainder of the code.

This is an example of my output:

  > summary(mod1)

       Call:
           glmmadmb(formula = species ~ (1 | year) + (1 | site), data = cs, 
           family = "nbinom2", link = "log")

        AIC: 8131.7 

       Coefficients:
        Estimate Std. Error z value Pr(>|z|)    
        (Intercept)     4.05       0.19    21.3   <2e-16 ***
          ---
         Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

       Number of observations: total=798, year=53, site=15 
       Random effect variance(s):
       Group=year
                  Variance StdDev
       (Intercept)   0.1925 0.4388
       Group=site
                  Variance StdDev
       (Intercept)   0.4748  0.689

       Negative binomial dispersion parameter: 1.7211 (std. err.: 0.088936)

       Log-likelihood: -4061.86 

I haven't found a function that extracts this value.

Thank you in advance!

Answer 1

The help page for glmmadmb , says, among other things:

Value:

  An object of class '"glmmadmb"' representing the model fit, including (among others) components: b: vector of fixed effects S: covariance matrix of random effects 

alpha: scale/overdispersion parameter (negative binomial, Gamma, beta)

So I think mod1$alpha (or mod1[["alpha"]] if you want to be very careful) should get what you want.

If the documentation weren't there, you could (1) look at the code of glmmADMB:::print.summary.glmmadmb as suggested by @DWin ; (2) look at names(mod1) , or str(mod1) , to find the piece of the model object that corresponded to the piece you wanted.

There probably should be an accessor method, but I don't know that there's a consistent convention in R for extracting dispersion-type parameters for models ...