predict_curve: Deprecated: Prediction curves for arbitrary functions of...

View source: R/predict_curve.R

predict_curveR Documentation

Deprecated: Prediction curves for arbitrary functions of posteriors

Description

Deprecated function for generating prediction curves (or a density for a prediction curve).

Usage

predict_curve(data, formula, summary = median, ...)

predict_curve_density(
  data,
  formula,
  summary = function(...) density_bins(..., n = n),
  n = 50,
  ...
)

Arguments

data

A data.frame, tibble, or grouped_df representing posteriors from a Bayesian model as might be obtained through spread_draws(). Grouped data frames as returned by group_by() are supported.

formula

A formula specifying the prediction curve. The left-hand side of the formula should be a name representing the name of the column that will hold the predicted response in the returned data frame. The right-hand side is an expression that may include numeric columns from data and variables passed into this function in ....

summary

The function to apply to summarize each predicted response. Useful functions (if you just want a curve) might be median(), mean(), or Mode(). If you want predictive distribution at each point on the curve, try density_bins() or histogram_bins().

...

Variables defining the curve. The right-hand side of formula is evaluated for every combination of values of variables in ....

n

For predict_curve_density, the number of bins to use to represent the distribution at each point on the curve.

Details

This function is deprecated. Use modelr::data_grid() combined with point_interval() or dplyr::do() and density_bins() instead.

The function generates a predictive curve given posterior draws (data), an expression (formula), and a set of variables defining the curve (...). For every group in data (if it is a grouped data frameā€”see group_by(); otherwise the entire data frame is taken at once), and for each combination of values in ..., the right-hand side of formula is evaluated and its results passed to the summary function. This allows a predictive curve to be generated, given (e.g.) some samples of coefficients in data and a set of predictors defining the space of the curve in ....

Given a summary function like median() or mean(), this function will produce the median (resp. mean) prediction at each point on the curve.

Given a summary function like density_bins(), this function will produce a predictive distribution for each point on the curve. predict_curve_density is a shorthand for such a call, with a convenient argument for adjusting the number of bins per point on the curve.

Value

If formula is in the form lhs ~ rhs and summary is a function that returns a single value, such as median or mode, then predict_curve returns a data.frame with a column for each group in data (if it was grouped), a column for each variable in ..., and a column named lhs with the value of summary(rhs) evaluated for every group in data and combination of variables in ....

If summary is a function that returns a data.frame, such as density_bins(), predict_curve has the same set of columns as above, except that in place of the lhs column is a set of columns named lhs.x for every column named x returned by summary. For example, density_bins() returns a data frame with the columns mid, lower, upper, and density, so the data frame returned by predict_curve with summary = density_bins will have columns lhs.mid, lhs.lower, lhs.upper, and lhs.density in place of lhs.

Author(s)

Matthew Kay

See Also

See density_bins().

Examples


# Deprecated; see examples for density_bins


tidybayes documentation built on Sept. 15, 2024, 9:08 a.m.