Skip to contents

Create or Test Objects of Class "mcpfamily"

Source: R/families.R
mcpfamily.Rd

[Experimental]

Usage

mcpfamily(x)

# S3 method for class 'mcpfamily'
format(x, ...)

# S3 method for class 'mcpfamily'
print(x, ...)

is.mcpfamily(x)

Arguments

x

A family object, e.g., binomial(link = "identity").

...

Must be empty. Reserved for future use.

Details

Converts standard R family objects into mcpfamily objects used internally by mcp. Supported family and link combinations include:

  • gaussian(link = "identity") or gaussian(link = "log")

  • binomial(link = "logit"), binomial(link = "probit"), or binomial(link = "identity")

  • bernoulli(link = "logit"), bernoulli(link = "probit"), or bernoulli(link = "identity")

  • poisson(link = "log") or poisson(link = "identity")

  • negbinomial(link = "log", link_shape = "log")

Note: mcpfamily objects are shipped with mcp - there is not (yet) support for user-supplied families.

Methods (by generic)

  • format(mcpfamily): Format an mcpfamily object.

  • print(mcpfamily): Print an mcpfamily object.

Functions

  • is.mcpfamily(): Checks whether x is an mcpfamily.

See also

family

Author

Jonas Kristoffer Lindeløv jonas@lindeloev.dk

Examples

# mcp() converts supported standard family objects automatically
data = data.frame(time = 1:6, y = exp(seq(0, 1, length.out = 6)))
my_normal = mcpfamily(stats::gaussian(link = "log"))
fit = mcp(list(y ~ 1), data, family = my_normal, par_x = "time", sample = FALSE)
family(fit)  # Show the mcp family retained in the fit
#> Family: gaussian
#> Links: mu = log; sigma = identity

# The converted object can also be inspected directly
mcpfamily(stats::binomial())$dpars  # Show its distributional parameters
#> [1] "mu"