How can I get logLik or AIC of a vglm fit

Question

I couldn't get them via normal methods.

> x <- rnorm(1000, 10, 1)
> y <-rgamma(1000, shape=1, rate=x)
> myfit <- vglm(y~x, family=gammaff(link="reciprocal"))
> summary(myfit)

Call:
vglm(formula = y ~ x, family = gammaff(link = "reciprocal"))

Pearson residuals:
         Min     1Q  Median      3Q     Max
1/mu -6.4126 -0.404 0.32118 0.69281 0.99917

Coefficients:
            Estimate Std. Error  z value
(Intercept) 0.050464     2.9415 0.017156
x           0.987358     0.2974 3.320015

Number of linear predictors:  1 

Name of linear predictor: 1/mu 

(Estimated) Dispersion Parameter for gammaff family:   0.96865

Residual deviance: 1106.494 on 998 degrees of freedom

Number of iterations: 6 
> AIC(myfit)
numeric(0)

> showMethods("AIC")
Function: AIC (package stats)
object="ANY"
object="cao"
object="qrrvglm"
object="rrvglm"
object="vgam"
object="vglm"
object="vlm"

Here it referes to AIC.vglm, but that doesn't even exist.

The closest function I found is AICrrvglm or AICvgam, neither works.

Answer 1

The problem seems to be that logLik.vlm(myfit) returns NULL. In the details of ?logLik.vlm, it states that

this code relies on the log-likelihood being defined, and computed, for the object.

In your case the log-likelihood is not computed: It is documented in ?"vglmff-class" about the loglikelihood:

loglikelihood: Object of class "function" returning the log-likelihood of the model. This slot is optional. If present, the function must have arguments function(mu, y, w, residuals = FALSE, eta, extra = NULL). The argument residuals can be ignored because log-likelihood residuals aren't defined.

Apparently gammaff does not contain this optional slot and therefore the loglikelihood was not computed and so the AIC cannot be computed either.