get_data_for_grid: Get raw data for plotting with model predictions

In mattansb/MSBMisc: Some functions I wrote that I find useful

Get raw data for plotting with model predictions

Description

Get raw data for plotting with model predictions

Usage

get_data_for_grid(grid, model, residualize = FALSE, collapse_by = FALSE, ...)

## S3 method for class 'data.frame'
get_data_for_grid(
  grid,
  model,
  residualize = FALSE,
  collapse_by = FALSE,
  pred_name,
  ...
)

## S3 method for class 'ggeffects'
get_data_for_grid(
  grid,
  model,
  residualize = FALSE,
  collapse_by = FALSE,
  protect_names = TRUE,
  ...
)

## S3 method for class 'emmGrid'
get_data_for_grid(
  grid,
  model,
  residualize = FALSE,
  collapse_by = FALSE,
  protect_names = TRUE,
  ...
)

## S3 method for class 'predictions'
get_data_for_grid(grid, model, residualize = FALSE, collapse_by = FALSE, ...)

Arguments

grid	A data grid with predictions
model	The statistical model
residualize	Should data be residualized?
collapse_by	Name of grouping variable to collaple across. If TRUE name of grouping variable is automatically detected from the model.
...	Args passed from / to other functions.
pred_name	Name of column that has the predictions in the data grid
protect_names	Logical, if TRUE, preserves column names from the ggeffects object.

Examples

data("mtcars")
mtcars <- mtcars |> transform(cyl = factor(cyl))
mod <- lm(mpg ~ hp + cyl, data = mtcars[1:10, ])

nd <- expand.grid(
  hp = seq(50, 350, by = 50),
  cyl = "4"
)

nd$predicted_mpg <- predict(mod, newdata = nd)

get_data_for_grid(nd, mod)

get_data_for_grid(nd, mod, residualize = TRUE, pred_name = "predicted_mpg")



library(ggplot2)
ggplot(nd, aes(hp, predicted_mpg)) +
  geom_line() +
  geom_point(aes(y = mpg, color = "Raw"),
    data = get_data_for_grid(nd, mod)
  ) +
  geom_point(aes(color = "Residualized"),
    data = get_data_for_grid(nd, mod, residualize = TRUE, pred_name = "predicted_mpg")
  ) +
  labs(
    title = "Partial residual plot",
    color = "Data"
  )

## Support of data-grid packages ------
# - ggeffects
# - emmeans
# - marginaleffects


pred_ggeffects <- ggeffects::ggpredict(mod, c("hp [50:350, by = 50]", "cyl [4]"))
get_data_for_grid(pred_ggeffects, residualize = TRUE)


at <- list(hp = seq(50, 350, by = 50), cyl = "4")
pred_emmeans <- emmeans::emmeans(mod, ~ hp + cyl, at = at)
get_data_for_grid(pred_emmeans, mod, residualize = TRUE)


# pred_marginaleffects <- marginaleffects::predictions(mod, newdata = nd)
# get_data_for_grid(pred_marginaleffects, residualize = TRUE)


## Collapes across group ------
fm1 <- lme4::lmer(angle ~ temperature + (1 | recipe),
  data = cake
)

pred_ggeffects <- ggeffects::ggpredict(fm1, c("temperature", "recipe"))
nd <- marginaleffects::datagrid(
  temperature = unique(cake$temperature),
  model = fm1
)
pred_marginaleffects <- marginaleffects::predictions(fm1, newdata = nd)

get_data_for_grid(pred_marginaleffects, collapse_by = TRUE)
# get_data_for_grid(pred_marginaleffects, collapse_by = TRUE, residualize = TRUE)

mattansb/MSBMisc documentation built on March 22, 2023, 6:02 p.m.