VGAM (version 1.1-9)

cops: Centre of the Parameter Space

Description

Returns a vector similar to coefs() comprising the centre of the parameter space (COPS) values, given a fitted VGLM regression.

Usage

cops(object, ...)
copsvglm(object, beta.range = c(-5, 6),
         tol = .Machine$double.eps^0.25,
         dointercepts = TRUE, trace. = FALSE,
         slowtrain = FALSE, ...)

Value

A named vector, similar to coefvlm. If trace. then a list is returned, having a componennt comprising a matrix of function evaluations used by

optimize.

Arguments

object

A vglm object. However, this function will not work for all such objects.

beta.range

Numeric. Interval for the numerical search. After a little scaling, it is effectively fed into interval in optimize. Convergence failure may occur if this argument is too wide so it is a good idea to vary this argument.

tol

Numeric. Fed into tol in optimize.

dointercepts

Logical. Compute the COPS for the intercepts? This should be set to FALSE for models such as propodds and cumulative.

trace.

Logical. Print a running log? This may or may not work properly.

slowtrain

Logical. If TRUE then all columns of a matrix is computed. If FALSE then only one column of a matrix is computed, and this is the only column needed.

...

currently unused but may be used in the future for further arguments passed into the other methods functions.

Author

Thomas W. Yee.

Details

For many models, some COPS values will be Inf or -Inf so that manual checking is needed, for example, poissonff. Each value returned may be effectively that of beta.range or NA. The answers returned by this function only make sense if the COPSs are in the interior of the parameter space.

References

Yee, T. W. (2021). Some new results concerning the Hauck-Donner effect. Manuscript in preparation.

See Also

hdeff. coefvlm, coef.

Examples

if (FALSE) data("xs.nz", package = "VGAMdata")
data1 <-
    na.omit(xs.nz[, c("age", "cancer", "sex")])
fit1 <- vglm(cancer ~ age + sex, binomialff,
             data = data1)
cops(fit1)