Influence diagnostics for a changepoint segmentation

Perturbs one observation at a time (deleting it, or replacing it with an outlier), re-runs the detector, and reports what changed: the number of changepoints, where they moved to, and how the segment parameters responded. This is the diagnostic family of Wilms, Killick and Matteson (2022), rendered in ggplot2 so it composes with the rest of the package.

Usage

cpt_influence(
  object,
  type = c("delete", "outlier"),
  engine = c("auto", "changepoint.influence", "recompute"),
  subset = NULL,
  outlier_sd = 5,
  seed = NULL,
  ...
)

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

# S3 method for class 'ggcpt_influence'
autoplot(
  object,
  plot_type = c("overview", "location", "parameter", "map"),
  ...
)

# S3 method for class 'ggcpt_influence'
tidy(x, ...)

Arguments

object

A ggcpt_influence object (for autoplot()).

type

"delete" (drop the observation) or "outlier" (replace it with a large value). Defaults to "delete".

engine

Which implementation to use: "auto" (default) uses changepoint.influence for a type = "delete" diagnostic of a change-in-mean fit from a changepoint engine when that package is installed (the only models it supports), and the generic recomputation otherwise; "changepoint.influence" insists on the former; "recompute" insists on the latter, which works for every wired and registered method.

subset

Optional integer vector of observation positions to perturb. Influence by deletion costs one detector fit per observation, so on a long series or an expensive engine this is the argument that makes the diagnostic affordable. Defaults to every observation.

outlier_sd

For type = "outlier", how many residual standard deviations the substituted value sits above the fitted level. Defaults to 5. Used by the recomputation engine, which is what "auto" picks for an outlier diagnostic; engine = "changepoint.influence" places its outliers by its own rule and does not read it.

seed

Optional seed, for detectors that randomise. The seed is scoped to this call: .Random.seed is saved and restored, so a seeded call inside a simulation loop does not pin the loop's own stream.

...

Additional arguments passed to cpt_detect() on each perturbed series.

x

A ggcpt_influence object (for print()).

plot_type

Which diagnostic to draw: "overview" (the series with the influential observations highlighted), "location" (perturbed changepoint locations against the perturbed observation), "parameter" (segment-parameter shift per perturbation) or "map" (the full influence map: perturbed observation on x, position on y, parameter shift as fill).

Value

A ggcpt_influence object: a list with

influence

a tibble with one row per perturbed observation: index, n_cp, delta_n_cp (against the unperturbed fit), max_shift (largest movement of a surviving changepoint, in positions), param_shift (largest absolute change in a segment parameter) and cpts (a list-column of the perturbed changepoint sets). A perturbation whose re-fit failed has n_cp = NA, and is warned about;

param

an \(n \times n\) matrix of per-observation segment parameters, one row per perturbation: the input to the influence map;

original, type, engine, method

with print() and autoplot() methods.

References

Wilms I, Killick R, Matteson DS (2022). “Graphical influence diagnostics for changepoint models.” Journal of Computational and Graphical Statistics, 31(3), 753–765. doi:10.1080/10618600.2021.2000873 .

See also

cpt_leverage(), cpt_sensitivity(), cpt_stability().

Examples

set.seed(2026)
fit <- cpt_detect(c(rnorm(40), rnorm(40, 4)), method = "pelt")
inf <- cpt_influence(fit)
inf
#> ggcpt_influence (delete perturbation, method: pelt, engine: changepoint.influence)
#>   Observations perturbed: 80
#>   Unperturbed changepoints: 1
#>   Perturbations changing the number of changepoints: 0 (0%)
#> 
#> Most influential observations:
#> # A tibble: 5 × 5
#>   index delta_n_cp max_shift param_shift leverage
#>   <int>      <int>     <dbl>       <dbl>    <dbl>
#> 1    40          0         1      0.0201     8.90
#> 2    15          0         0      0.0642     2.93
#> 3     6          0         0      0.0635     2.88
#> 4    77          0         0      0.0578     2.50
#> 5    26          0         0      0.0500     1.97
head(cpt_leverage(inf))
#> # A tibble: 6 × 5
#>   index delta_n_cp max_shift param_shift leverage
#>   <int>      <int>     <dbl>       <dbl>    <dbl>
#> 1    40          0         1      0.0201     8.90
#> 2    15          0         0      0.0642     2.93
#> 3     6          0         0      0.0635     2.88
#> 4    77          0         0      0.0578     2.50
#> 5    26          0         0      0.0500     1.97
#> 6    27          0         0      0.0454     1.66