Skip to contents

Print fit summaries and diagnostics, or plot stored common-grid predictions and their differences from the continuous benchmark.

Usage

# S3 method for class 'bin_interaction'
print(x, digits = 4L, ...)
# S3 method for class 'bin_interaction'
plot(x, type = c("predictions", "differences"),
  main = NULL, xlab = NULL, ylab = NULL,
  show_confidence = type == "predictions",
  level = if (is.null(x$level)) 0.95 else x$level, ...)

Arguments

x

A bin_interaction result.

digits

Significant digits in printed numeric summaries.

type

Plot predictions or differences from the continuous benchmark.

main, xlab, ylab

Optional overall title and axis labels.

show_confidence

Show shaded pointwise confidence intervals for the mean. Defaults to TRUE for predictions and FALSE for differences. TRUE with a difference plot is an error because paired difference intervals are not implemented.

level

Confidence level strictly between 0 and 1. Defaults to the saved level, or 0.95 for older objects. Changing it does not refit the models.

...

Unused by print. For plot, additional base plot arguments, excluding supplied limits, labels, title, or plotting coordinates.

Details

The horizontal axis is the first focal variable in original units. With a second focal variable, panels use its stored conditioning values, with common horizontal and vertical scales. Controls remain at the stored reference values. Colors and line types identify models. Each focal bin is drawn separately to avoid connecting across a categorical discontinuity; a single supported point is marked as a dot. Unavailable predictions leave gaps. Failed or warning fits are not plotted. Logit predictions use a 0-to-1 probability scale. Difference plots show candidate predictions minus the continuous benchmark; they do not report statistical significance. The legend appears in the first panel. Use a larger device or RStudio Zoom for the three-panel interaction view.

Prediction plots show confidence bands by default. Bands are computed from the saved models and drawn separately within each focal bin, with a vertical interval at an isolated point such as a separate zero category. The caption states the confidence level. See predict.bin_interaction for the model-based assumptions. These pointwise intervals condition on fixed model settings, including cutpoints and any supplied spline complexity. Their overlap or non-overlap is not a formal test of a difference between models fitted to the same observations. Difference plots retain point estimates only. Graphics parameters are restored after plotting.

Value

The input object invisibly. Plots use the active graphics device.