Define the transformation applied to observed data

mc_observation_transformation() defines the necessary transformation applied to the observed data before passing it to generate model check.

mc_observation_transformation(transform = NULL, group = NULL)

Arguments

transform

The transform function that is applied on the response variable in observed data. Default to be NULL. If NULL, vmc will use no transformation. The transform function takes an input of a data frame, e.g., the data frame passed to observation, containing a column named observation standing for observations of the response variable and several columns standing for the predictors in model (if any). The output of the transform function should be in the same form as the input, a data frame containing a column named observation and several columns for the predictors. This argument is useful when mc_draw is set to a distribution that is in a different unit from the raw observation, e.g., sigma in Gaussian family models describes the variance of observation. See example for more details.

group

A set of variables quoted by ggplot2::vars() and represents the group option used before applying the transform function.

Details

The transformation is necessary when the distribution generated by the model is not predictive distributions. For example, when the distribution generated from phi parameter of beta distribution, the observed data can not be shown in model check directly but need to be transformed into the same scale of phi.

Examples

library(ggplot2)

mcplot(mpg_model) +
  mc_draw("sigma") +
  mc_observation_transformation(sd, vars(disp)) +
  mc_gglayer(coord_flip())
#> Warning: Removed 23 rows containing non-finite outside the scale range
#> (`stat_density()`).
#> Warning: No shared levels found between `names(values)` of the manual scale and the
#> data's fill values.