posterior_epred_ordbeta.brmsfit

Calculate Probability of Response Components

Description

This function is an alternative to the brms default posterior_epred to allow for predictions of the probability of the bottom, top, or middle (i.e. continuous) parts of the response. Useful when wanting to understand what the effect of a covariate is on bottom or top values of the scale.

Usage

## S3 method for class 'brmsfit'
posterior_epred_ordbeta(
  object,
  component = "all",
  newdata = NULL,
  re_formula = NULL,
  re.form = NULL,
  resp = NULL,
  dpar = NULL,
  nlpar = NULL,
  ndraws = NULL,
  draw_ids = NULL,
  sort = FALSE,
  ...
)

Arguments

`object`	An ordbetareg/brms object
`component`	The type of response component, i.e., the probability of the bottom end of the scale, the top end, or the middle (i.e.) continuous values.
`newdata`	see brms::posterior_epred
`re_formula`	see brms::posterior_epred
`re.form`	see brms::posterior_epred
`resp`	see brms::posterior_epred
`dpar`	see brms::posterior_epred
`nlpar`	see brms::posterior_epred
`ndraws`	see brms::posterior_epred
`draw_ids`	see brms::posterior_epred
`sort`	see brms::posterior_epred
`…`	see brms::posterior_epred

Details

To predict the top, bottom, or "middle" (i.e. continuous) components of the response, set the component argument to "top", "bottom" or "continuous". By default, component is set to "all", which will replicate behavior of the default posterior_epred function.

All other arguments besides component are the same as the standard generic posterior_predict. For more information on the relevant arguments for posterior_epred, see brms::posterior_epred.

Value

An S x N matrix where S is the number of posterior draws and N is the number of observations.

Examples

library(ordbetareg)


data('ord_fit_mean')

# use function to calculate probability of top end of scale

pr_1s <- posterior_epred_ordbeta(ord_fit_mean,component="top")

# use function to calculate probability of bottom end of scale

pr_0s <- posterior_epred_ordbeta(ord_fit_mean,component="top")

# use function to calculate probability of continuous /
# beta-distributed part of scale

pr_beta <- posterior_epred_ordbeta(ord_fit_mean,component="top")