cpt_segment_models: Fit a model to every segment of a segmentation

In ggchangepoint: Combines Changepoint Analysis with 'ggplot2'

Fit a model to every segment of a segmentation

Description

Detection says where the regimes change; this fits your model to each regime. Every segment of fit gets its own copy of model, fitted by engine to that segment's rows, and the result is a table with one fitted model per segment that tidy(), glance() and predict() read.

Usage

cpt_segment_models(fit, model = NULL, data = NULL, engine = stats::lm, ...)

## S3 method for class 'ggcpt_segment_models'
tidy(x, conf_level = 0.95, ...)

## S3 method for class 'ggcpt_segment_models'
glance(x, ...)

## S3 method for class 'ggcpt_segment_models'
print(x, ...)

Arguments

fit	A ggcpt object.
model	A formula. For a result from a formula fit it defaults to that formula; for a series it defaults to value ~ 1 (the segment level), or value ~ time for a change in slope. The series is available as value, its position as time, and the time index (when the result carries one) as index.
data	Optional data frame with one row per observation, supplying covariates the model uses. For a formula fit the covariates are taken from the result.
engine	The fitting function, called as engine(model, data = segment_rows, ...). Defaults to stats::lm; stats::glm with a family works the same way.
...	Further arguments for engine.
x	A ggcpt_segment_models object.
conf_level	Confidence level for the coefficient intervals.

Value

A ggcpt_segment_models tibble with one row per segment: segment, start, end, n and model (a list-column of fitted models, NULL where the fit failed). tidy() gives the coefficient table with a segment column; glance() one row per segment with n, sigma, r_squared, AIC, BIC and logLik.

See Also

predict.ggcpt() to forecast from a segment, tidy.ggcpt(what = "coefficients") for the coefficients the detection itself implies.

Other segment models: cpt_counterfactual(), predict.ggcpt()

Examples

set.seed(1)
d <- data.frame(x = runif(200))
d$y <- ifelse(seq_len(200) <= 100, 1 + 2 * d$x, 3 - d$x) + rnorm(200, 0, 0.3)
fit <- cpt_detect(y ~ x, data = d, method = "strucchange")
sm <- cpt_segment_models(fit)
tidy(sm)
glance(sm)

ggchangepoint documentation built on Nov. 5, 2026, 9:06 a.m.