its2s

its2s /ˈɪts.tuː.ɛs/ noun

An open-source Python package for interrupted time series, 2-stage counterfactual analysis with moving-block bootstrap confidence intervals.
A modular framework for fitting ITS models with cross-validation, hyperparameter tuning, and reproducible end-to-end workflows.

Why `its2s`?¶

Flexibility: its2s is designed to be flexible, allowing you to easily use this methodology with any data, model, and settings.
Modularity: its2s is designed to be modular, allowing you to use the parts of the pipeline that you need.
Reproducibility: its2s is designed to be reproducible via seeding and rigorous parameter documentation, allowing you to reproduce the results of your analysis easily.
Efficiency: its2s is designed to run quickly and efficiently, allowing you to run the analysis quickly and easily.

Function	Overview	Inputs	Outputs
`run_single_its`	End-to-end ITS pipeline for a single series: split, fit, bootstrap, score, save outputs.	`df`, `intervention_date`, `model_name` (`prophet_xgb` / `prophet_then_xgb` / `neuralprophet` / `arima`), optional `config_path` / `config_overrides`, optional `output_dir`	`PipelineResult` (+ files when `output_dir` is set)
`run_batch`	Run the pipeline over many series, optionally in parallel.	`series_list` (list of dicts with `series_id`, `df`, `intervention_date`, …), optional `config_path`, `output_dir`, `n_jobs`, `seed`	list of `PipelineResult`

from its2s import run_single_its, run_batch

Function	Overview	Inputs	Outputs
`compare_models`	Fit several models on the same series and compare metrics side-by-side.	`df`, `intervention_date`, `model_names`, optional config / overrides, `output_dir`	comparison DataFrame + per-model `PipelineResult`
`tune_model`	Latin hypercube hyperparameter search via expanding-window CV. Run before `run_single_its` to pick hyperparameters, then pass `best_params` via `config_overrides`.	`df`, `intervention_date`, `model_name`, `n_trials`, `n_folds`, optional `metric` / `cv_end_date`	`TuningResult` (best params + per-trial scores)

from its2s import compare_models, tune_model

Setup — install the package into a Python environment.
Quick Start — a minimal example using simulated data, plus your-own-data, outputs, covariates, and configuration.
Tutorials — six step-by-step notebooks: data splitting, cross-validation, hyperparameter tuning, model variants, moving-block bootstrap, and a full end-to-end workflow.
API Reference — function signatures and docstrings for the public API.
Citation — how to cite its2s.

If you have a question, feature request, or bug, please open an issue.