summary.posterior_diff: Summarize Posterior Distributions of Model Differences

Description Usage Arguments Details Value Examples

Description

Credible intervals are created for the differences. Also, region of practical equivalence (ROPE) statistics are computed when the effective size of a difference is given.

Usage

1
2
## S3 method for class 'posterior_diff'
summary(object, prob = 0.9, size = 0, ...)

Arguments

object

An object produced by contrast_models().

prob

A number p (0 < p < 1) indicating the desired probability mass to include in the intervals.

size

The size of an effective difference. For example, a 5% increase in accuracy between two models might be considered a "real" difference.

...

Not currently used

Details

The ROPE estimates included in the results are the columns pract_neg, pract_equiv, and pract_pos. pract_neg integrates the portion of the posterior below -size (and pract_pos is the upper integral starting at size). The interpretation depends on whether the metric being analyzed is better when larger or smaller. pract_equiv integrates between [-size, size]. If this is close to one, the two models are unlikely to be practically different relative to size.

Value

A data frame with interval and ROPE statistics for each comparison.

Examples

1
2
3
4
5
6
7
8
9
# Example objects from the "Getting Started" vignette at
#  https://topepo.github.io/tidyposterior/articles/Getting_Started.html

# File for pre-run model is at
ex_dat <- "https://bit.ly/2S1v6H9"

# load(load(url(ex_dat))

# summary(glm_v_nnet, size = 0.02)

tidyposterior documentation built on May 2, 2019, 2:49 p.m.