Link Functions for VGLMs

Returns the link functions, and parameter names, for vector generalized linear models (VGLMs).

linkfun(object, ...)
linkfunvlm(object, earg = FALSE, by.var = FALSE, ...)

Arguments

object

An object which has parameter link functions, e.g., has class "vglm".

earg

Logical. Return the extra arguments associated with each link function? If TRUE then a list is returned.

by.var

Logical. If TRUE then a named character vector is returned, corresponding to coef(object), with the link function corresponding to each regression coefficient.

...

Arguments that might be used in the future.

Details

All fitted VGLMs have a link function applied to each parameter. This function returns these, and optionally, the extra arguments associated with them.

Value

By default, usually just a (named) character string, with the link functions in order. It is named with the parameter names. If earg = TRUE then a list with the following components.

link

The default output.

earg

The extra arguments, in order.

If by.var = TRUE then what is returned is described above.

Author

Thomas W. Yee

Note

Presently, the multinomial logit model has only one link function, multilogitlink, so a warning is not issued for that link. For other models, if the number of link functions does not equal \(M\) then a warning may be issued.

See also

linkfun, multilogitlink, vglm, coefvlm.

Examples

pneumo <- transform(pneumo, let = log(exposure.time))
fit1 <- vglm(cbind(normal, mild, severe) ~ let, propodds, data = pneumo)
coef(fit1, matrix = TRUE)
#>             logitlink(P[Y>=2]) logitlink(P[Y>=3])
#> (Intercept)          -9.676093         -10.581725
#> let                   2.596807           2.596807
linkfun(fit1)
#>     P[Y>=2]     P[Y>=3] 
#> "logitlink" "logitlink" 
linkfun(fit1, earg = TRUE)
#> $link
#>     P[Y>=2]     P[Y>=3] 
#> "logitlink" "logitlink" 
#> 
#> $earg
#> $earg$`P[Y>=2]`
#> $earg$`P[Y>=2]`$theta
#> 
#> 
#> $earg$`P[Y>=2]`$bvalue
#> NULL
#> 
#> $earg$`P[Y>=2]`$inverse
#> [1] FALSE
#> 
#> $earg$`P[Y>=2]`$deriv
#> [1] 0
#> 
#> $earg$`P[Y>=2]`$short
#> [1] TRUE
#> 
#> $earg$`P[Y>=2]`$tag
#> [1] FALSE
#> 
#> attr(,"function.name")
#> [1] "logitlink"
#> 
#> $earg$`P[Y>=3]`
#> $earg$`P[Y>=3]`$theta
#> 
#> 
#> $earg$`P[Y>=3]`$bvalue
#> NULL
#> 
#> $earg$`P[Y>=3]`$inverse
#> [1] FALSE
#> 
#> $earg$`P[Y>=3]`$deriv
#> [1] 0
#> 
#> $earg$`P[Y>=3]`$short
#> [1] TRUE
#> 
#> $earg$`P[Y>=3]`$tag
#> [1] FALSE
#> 
#> attr(,"function.name")
#> [1] "logitlink"
#> 
#> 

fit2 <- vglm(cbind(normal, mild, severe) ~ let, multinomial, data = pneumo)
coef(fit2, matrix = TRUE)
#>             log(mu[,1]/mu[,3]) log(mu[,2]/mu[,3])
#> (Intercept)          11.975092          3.0390622
#> let                  -3.067466         -0.9020936
linkfun(fit2)
#> [1] "multilogitlink"
linkfun(fit2, earg = TRUE)
#> $link
#> [1] "multilogitlink"
#> 
#> $earg
#> $earg$multilogitlink
#> $earg$multilogitlink$M
#> [1] 2
#> 
#> $earg$multilogitlink$refLevel
#> [1] 3
#> 
#> 
#>