Package index • ggchangepoint

Core API

cpt_detect(): Unified changepoint detection dispatcher
cpt_methods(): Introspect available changepoint detection methods
cpt_families(): Distribution families, and what each request runs
cpt_penalty(): Construct changepoint penalties
cpt_cite(): Cite the method behind a result
as_cpt_series(): Coerce a time series object to values plus a time index
new_ggcpt(): Create a ggcpt object
is_ggcpt(): Test if an object is a ggcpt object
print(<ggcpt>): Print a ggcpt object
cpt_verify() print(<ggcpt_verification>) tidy(<ggcpt_verification>): Re-run a result against the engines installed now

The contract

What code that calls the package can rely on: the classed conditions it raises, and the versioned JSON schema a result is written in.

ggchangepoint-conditions ggchangepoint_error ggchangepoint_warning: Conditions raised by ggchangepoint
as_json(): Convert a result to JSON
cpt_export() cpt_import(): Write a result to a file, and read it back

Extending the package

Bring a detector this package does not wrap (a non-CRAN engine, a Python tool, a neural detector, or your own changepoints) into the same tidy, plottable grammar.

as_ggcpt(): Turn external changepoints into a ggcpt result
cpt_register_method() cpt_unregister_method() cpt_registered_methods(): Register an external changepoint detector
cpt_install_engines(): Install the engines behind a family of methods

Original wrappers

cpt_wrapper(): Changepoint wrapper
ecp_wrapper(): ecp wrapper

Search and pruning wrappers

fpop_wrapper(): FPOP wrapper: Functional Pruning Optimal Partitioning
wbs_wrapper(): WBS wrapper: Wild Binary Segmentation
wbs2_wrapper(): WBS2 wrapper: Wild Binary Segmentation 2
not_wrapper(): NOT wrapper: Narrowest-Over-Threshold
mosum_wrapper(): MOSUM wrapper: Moving Sum
idetect_wrapper(): Isolate-Detect wrapper
tguh_wrapper(): TGUH wrapper

Inference and slope wrappers

smuce_wrapper(): SMUCE / HSMUCE wrapper: multiscale changepoint inference
cpop_wrapper(): CPOP wrapper: optimal change-in-slope detection
nsp_wrapper(): NSP wrapper: Narrowest Significance Pursuit
cpt_crops() autoplot(<ggcpt_path>) print(<ggcpt_path>) tidy(<ggcpt_path>): CROPS: the full penalty path of a penalised changepoint method

Inference

Where could this changepoint be, is it real, and which coordinates changed?

cpt_confint(): Confidence intervals for changepoint locations
cpt_confint_coverage() print(<ggcpt_coverage>) tidy(<ggcpt_coverage>) autoplot(<ggcpt_coverage>): How often do the confidence intervals cover? A Monte Carlo check
cpt_test(): Test detected changepoints
cpt_regions(): Tidy the significance regions of a ggcpt object
cpt_attribute() print(<ggcpt_attribution>) tidy(<ggcpt_attribution>) glance(<ggcpt_attribution>) autoplot(<ggcpt_attribution>): Which coordinates changed? Certified attribution for a multivariate fit
cpt_screen() print(<ggcpt_screen>) tidy(<ggcpt_screen>) glance(<ggcpt_screen>) autoplot(<ggcpt_screen>): Which of many series changed? Screening with error control across the panel

Effects, dates and empty answers

How big was the change, did anything change at a date fixed in advance, was the change the event you know about, and what does “no changepoints” mean.

cpt_effect() print(<ggcpt_effect>) tidy(<ggcpt_effect>) autoplot(<ggcpt_effect>): Effect size at each changepoint
cpt_test_at() print(<ggcpt_test_at>) tidy(<ggcpt_test_at>): Test for a change at a date fixed in advance
cpt_attribute_event(): Was a detected change the event you have in mind?
cpt_test_null(): Test for a change anywhere in a series
cpt_null_power() print(<ggcpt_null_power>) tidy(<ggcpt_null_power>): What an empty answer could have detected

Segment models

A model per regime, a forecast from the last one, and what would have happened without each change.

cpt_segment_models() tidy(<ggcpt_segment_models>) glance(<ggcpt_segment_models>) print(<ggcpt_segment_models>): Fit a model to every segment of a segmentation
predict(<ggcpt>): Forecast from a segment of a segmentation
cpt_counterfactual() print(<ggcpt_counterfactual>) tidy(<ggcpt_counterfactual>) glance(<ggcpt_counterfactual>) autoplot(<ggcpt_counterfactual>): What would have happened without the change? Observed against counterfactual

Choosing the number of changepoints

cpt_select() print(<ggcpt_selection>) tidy(<ggcpt_selection>) autoplot(<ggcpt_selection>): Choose the number of changepoints
cpt_k_weights() print(<ggcpt_k_weights>) tidy(<ggcpt_k_weights>) glance(<ggcpt_k_weights>) autoplot(<ggcpt_k_weights>): How sure is the number of changepoints? Weights over K

Diagnostics

Which observation is driving this, which setting, and what did the detector actually compute?

cpt_influence() print(<ggcpt_influence>) autoplot(<ggcpt_influence>) tidy(<ggcpt_influence>): Influence diagnostics for a changepoint segmentation
cpt_leverage(): Rank observations by influence
cpt_sensitivity() print(<ggcpt_sensitivity>) tidy(<ggcpt_sensitivity>) autoplot(<ggcpt_sensitivity>): Sensitivity of a segmentation to its tuning parameters
cpt_statistic() ggcpt_statistic(): The detector's statistic as a function of location
cpt_solution_path() ggcpt_solution_path(): The solution path of a search-based detector
cpt_scale_space() ggcpt_scale_space(): Scale space: the statistic across bandwidths

Assumptions and robustness

Do the result’s assumptions hold for this series, and does the answer survive a change of noise model?

cpt_assumptions() print(<ggcpt_assumptions>) tidy(<ggcpt_assumptions>): Check the assumptions behind a segmentation
cpt_gof() print(<ggcpt_gof>) tidy(<ggcpt_gof>): Goodness of fit, segment by segment
cpt_abruptness() print(<ggcpt_abruptness>) tidy(<ggcpt_abruptness>) autoplot(<ggcpt_abruptness>): Was the change abrupt? A step against a ramp, at every changepoint
cpt_robustness() print(<ggcpt_robustness>) tidy(<ggcpt_robustness>) autoplot(<ggcpt_robustness>): Does a changepoint survive a change of noise model?

Supervised detection

Labelled regions as ground truth, label errors as the accuracy measure, and a learned penalty.

cpt_labels() tidy(<cpt_labels>): Changepoint labels
as_cpt_labels(): Coerce annotations to changepoint labels
cpt_label_error() tidy(<cpt_label_error>) print(<cpt_label_error>): Score a segmentation against labels
cpt_label_error_curve() print(<ggcpt_label_curve>) tidy(<ggcpt_label_curve>) autoplot(<ggcpt_label_curve>): Label error as a function of the penalty
cpt_learn_penalty() print(<ggcpt_penalty_model>) coef(<ggcpt_penalty_model>) tidy(<ggcpt_penalty_model>) predict(<ggcpt_penalty_model>): Learn a penalty from labelled series

Choosing and combining methods

cpt_recommend() tidy(<ggcpt_recommendation>) print(<ggcpt_recommendation>) autoplot(<ggcpt_recommendation>): Recommend a detection method
cpt_consensus() autoplot(<ggcpt_consensus>) print(<ggcpt_consensus>): Consensus changepoints across several detectors

Communication

cpt_annotate_events() print(<ggcpt_events>) tidy(<ggcpt_events>) autoplot(<ggcpt_events>): Match detected changepoints to known events
cpt_report(): A reproducible report of a changepoint analysis
cpt_gt(): A publication-ready changepoint table

Benchmarking

cpt_benchmark() print(<ggcpt_benchmark>) tidy(<ggcpt_benchmark>) autoplot(<ggcpt_benchmark>): Benchmark detectors across datasets
cpt_datasets(): A catalogue of benchmark datasets
cpt_load_tcpd(): Download and cache the Turing Change Point Dataset
cpt_annotations(): Per-annotator ground truth for a benchmark dataset

Streaming and online monitoring

cpt_monitor() tidy(<ggcpt_monitor>) print(<ggcpt_monitor>) autoplot(<ggcpt_monitor>): A stateful sequential changepoint monitor
cpt_update(): Feed observations to a monitor
alarms(): The alarm log of a monitor
cpt_replay(): Replay a series through a sequential detector
cpt_delay() tidy(<ggcpt_delay>) glance(<ggcpt_delay>) print(<ggcpt_delay>) autoplot(<ggcpt_delay>): Detection delay and false-alarm rate

Power and study design

cpt_power() tidy(<ggcpt_power>) print(<ggcpt_power>) autoplot(<ggcpt_power>): Detection power for a changepoint scenario
cpt_min_detectable() tidy(<ggcpt_min_detectable>) print(<ggcpt_min_detectable>): The smallest detectable change
cpt_scenarios(): A grid of simulation scenarios

Bayesian wrappers

bcp_wrapper(): Bayesian changepoint wrapper (Barry-Hartigan product partition model)
bocpd_wrapper(): Bayesian online changepoint detection wrapper (BOCPD)
beast_wrapper(): BEAST wrapper: Bayesian estimation of abrupt change, seasonality, and trend

Nonparametric and sequential wrappers

cpm_wrapper(): Sequential change point model wrapper (CPM)
kcp_wrapper(): Kernel changepoint wrapper (KCP on running statistics)
npmojo_wrapper(): Nonparametric MOSUM wrapper (NP-MOJO)
sn_wrapper(): Self-normalisation wrapper (SNSeg)

Robust and model-selection wrappers

decafs_wrapper(): DeCAFS wrapper: changes amid drift and autocorrelated noise
envcpt_wrapper(): EnvCpt wrapper: changepoints versus trends versus autocorrelation
fastcpd_wrapper(): fastcpd wrapper: fast changepoint detection via sequential gradient descent

Multivariate and high-dimensional wrappers

inspect_wrapper(): inspect wrapper: high-dimensional changepoints via sparse projection
ocd_wrapper(): ocd wrapper: online high-dimensional changepoint detection
geomcp_wrapper(): Geometrically-inspired multivariate changepoint wrapper (geomcp)
esac_wrapper(): ESAC wrapper: sparsity-adaptive high-dimensional detection
pilliat_wrapper(): Pilliat wrapper: high-dimensional detection by three complementary tests
hdcov_wrapper(): High-dimensional covariance changepoints
network_wrapper(): Dynamic-network changepoints
var_wrapper(): VAR(1) changepoints
hdreg_wrapper(): High-dimensional regression changepoints

Functional and network wrappers

fmean_wrapper(): Functional mean changepoints
fcov_wrapper(): Functional covariance changepoints
kwc_wrapper(): Robust depth-based changepoints for functional and multivariate data
fabisearch_wrapper(): Network-structure changepoints via non-negative matrix factorisation

Regression-break wrappers

strucchange_wrapper(): Bai-Perron structural break wrapper (strucchange)
segmented_wrapper(): Broken-line regression wrapper (segmented)
bfast_wrapper(): BFAST wrapper: breaks for additive season and trend

Applied vocabularies and fast paths

trend_wrapper(): Classical single-changepoint tests (Pettitt, Buishand, SNHT)
taylor_wrapper(): Taylor's change point analyzer
wbsts_wrapper(): WBS for nonstationary time series
binsegrcpp_wrapper(): Fast binary segmentation across loss functions

Epidemic changes and anomalies

Stretches that leave a baseline and return to it, and single outliers.

capa_wrapper(): CAPA wrapper: collective and point anomalies (epidemic changepoints)

Bayesian formula models

mcp_wrapper(): Bayesian formula-based changepoint regression (mcp)

broom methods

tidy(<ggcpt>): Tidy a ggcpt object
glance(<ggcpt>): Glance at a ggcpt object
augment(<ggcpt>): Augment a ggcpt object
summary(<ggcpt>) print(<summary.ggcpt>): Summary of a ggcpt object

Additional S3 methods

as_tibble(<ggcpt>) as.data.frame(<ggcpt>) format(<ggcpt>) plot(<ggcpt>): Coerce, format, and plot ggcpt objects
plot(<ggcpt_selection>) plot(<ggcpt_stability>) plot(<ggcpt_sensitivity>) plot(<ggcpt_influence>) plot(<ggcpt_batch>) plot(<ggcpt_benchmark>) plot(<ggcpt_consensus>) plot(<ggcpt_monitor>) plot(<ggcpt_delay>) plot(<ggcpt_path>) plot(<ggcpt_power>) plot(<ggcpt_effect>) plot(<ggcpt_robustness>) plot(<ggcpt_recommendation>) plot(<ggcpt_events>) plot(<ggcpt_label_curve>) plot(<ggcpt_attribution>) plot(<ggcpt_coverage>) plot(<ggcpt_counterfactual>) plot(<ggcpt_abruptness>) plot(<ggcpt_screen>) plot(<ggcpt_k_weights>): Base plot() methods for ggchangepoint result objects

Theming, palettes and accessibility

theme_ggcpt(): ggchangepoint theme
annotate_segments(): Annotate segments with alternating shading
scale_colour_cpt() scale_color_cpt() scale_fill_cpt() scale_linetype_cpt(): Colour-vision-safe scales for changepoint methods
scale_fill_cpt_label() scale_colour_cpt_label(): Colour scales for changepoint labels and label errors

Visualization

autoplot(<ggcpt>): Autoplot a ggcpt object
ggcptplot(): Plot for the changepoint package
ggecpplot(): Plot for the ecp package
geom_changepoint(): Changepoint vertical rules geom
geom_cpt_segment(): Changepoint segment level geom
geom_cpt_ci(): Changepoint confidence interval geom
geom_cpt_region(): Significance region geom
geom_cpt_label(): Changepoint label geom
geom_cpt_event(): Event annotation geom
stat_changepoint(): Changepoint detection stat
stat_cpt_region(): Significance regions computed in the layer
ggchangepoint-ggproto GeomChangepoint GeomCptSegment GeomCptCi GeomCptRegion GeomCptLabel GeomCptEvent StatChangepoint StatCptRegion: ggproto objects for the changepoint layers
ggcpt_posterior(): Posterior probability plot for Bayesian results
ggcpt_runlength(): Run-length posterior heatmap for Bayesian online results
ggcpt_interactive(): Interactive changepoint plot

Method comparison, batch, and stability

ggcpt_compare(): Compare multiple changepoint detection methods
ggcpt_compare_table(): Comparison table
cpt_batch() print(<ggcpt_batch>) tidy(<ggcpt_batch>) autoplot(<ggcpt_batch>): Batch changepoint detection over many series
cpt_stability() print(<ggcpt_stability>) tidy(<ggcpt_stability>) autoplot(<ggcpt_stability>): Changepoint stability diagnostics via bootstrap

Evaluation

cpt_metrics(): Changepoint accuracy metrics
cpt_metrics_annotated(): Multi-annotator evaluation
ggcpt_eval(): Evaluation visualization

Simulation and data

cpt_simulate() rcpt(): Generate simulated changepoint data
signal_blocks(): Blocks test signal
signal_fms(): FMS (Four-Metric-Segments) test signal
signal_mix(): Mix test signal
signal_teeth(): Teeth test signal
signal_stairs(): Stairs test signal

Measurements

What the engines were measured to do at this release: runtimes, invariances, false positives under four noise regimes and four data types, and false alarms on pure noise. cpt_methods() and cpt_recommend() read them.

cpt_runtimes: Engine runtimes by series length
cpt_invariances: Invariances of every engine's answer
cpt_noise_benchmark: False positives and power under four noise regimes
cpt_data_types: False positives and power by data type
cpt_null_sizes: False alarms on pure noise at scale
cpt_calibration: What the package promises, against what it delivers

Re-exports

reexports tidy glance augment autoplot as_tibble: Objects exported from other packages