df_for_wald {glmglrt}

Gets the degree of freedom for Wald tests involving the model

Description

This generic function is used by p_value_contrast.default to get the number of degrees of freedom of the t distribution that approximates the point estimate of the contrast divided by its standard error.

Usage

df_for_wald(object, ...)

## S3 method for class 'glm'
df_for_wald(object, ...)

## Default S3 method:
df_for_wald(object, ...)

Arguments

object

statistical model;

...

Unused by p_value_contrast.default, but may be useful to some custom specializations.

Details

This function is quite similar to df.residual but it should return Inf when the Student's t distribution is less appropriate than the normal distribution.

Value

A finite value or Inf for normal distribution approximation.

Methods (by class)

• glm: Returns df.residual for linear gaussian models and Inf for all other models in order to make Wald's tests consistent with the behavior of stats::summary.glm(object)

• default: Simple proxy to df.residual but replaces NAs with Inf

See Also

Other Wald-related functions: fixcoef(), p_value_contrast(), vcov_fixcoef()

Examples

# 10 observations, one coefficient, 9 degrees of freedom
df_for_wald(glm(I(1:10) ~ 1))
# returns Inf (non-gaussian-identity model)
df_for_wald(glm(family="poisson", c(10,20,30) ~ 1))
data(mtcars)
# returns Inf (non-gaussian-identity model)
df_for_wald(glm(family="binomial", data=mtcars, I(hp > median(hp)) ~ cyl))