# greybox

> Toolbox for regression model building, selection, and forecasting
> evaluation, in R (CRAN) and Python (PyPI). Core estimator: ALM
> (Augmented Linear Model), supporting 26 distributions and 7 loss
> functions, plus stepwise selection, model combination, rolling-origin
> cross-validation, accuracy measures, and exploratory decomposition.
> Source: https://github.com/config-i1/greybox.
> Statistics for Business Analytics book: https://openforecast.org/sba/.

Status legend used below: `stable` = implemented and in regular use in
both R and Python; `R-only` = R has it, Python port pending; `partial` =
exists in Python but missing features compared to R.

## Navigation

- [Home](https://github.com/config-i1/greybox/wiki/Home): wiki landing page with the function index and the R-Python parity matrix.
- [Glossary](https://github.com/config-i1/greybox/wiki/Glossary): disambiguation of overloaded terms (distribution values vs d/p/q/r functions, loss vs ic, lags meanings, R `$` vs Python `.`).
- [Roadmap](https://github.com/config-i1/greybox/wiki/Roadmap): functionality that exists in R greybox but is not yet in the Python port.
- [R-Python-differences](https://github.com/config-i1/greybox/wiki/R-Python-differences): numerical-parity status and intentional convention/interface differences.
- [Installation](https://github.com/config-i1/greybox/wiki/Installation): install instructions for R (CRAN) and Python (PyPI).
- [Resources](https://github.com/config-i1/greybox/wiki/Resources): academic references and DOIs.

## Core estimator

- [ALM](https://github.com/config-i1/greybox/wiki/ALM) — `alm()` (R) / `ALM` (Python). Stable. Augmented Linear Model: 26 distributions, 7 loss functions.
- [CALM](https://github.com/config-i1/greybox/wiki/CALM) — `calm()` / `CALM`. Stable. Combination of ALMs (model averaging by information criteria).
- [stepwise](https://github.com/config-i1/greybox/wiki/stepwise) — `stepwise()` / `stepwise`. Stable. Forward stepwise variable selection.
- [distributions](https://github.com/config-i1/greybox/wiki/distributions) — `d/p/q/r` functions. Stable. Distribution families used by ALM.

## Exploratory data analysis

- [EDA](https://github.com/config-i1/greybox/wiki/EDA) — `stick()` (R and Python). Stable. Seasonality/Trend/Irregular (STI) variance decomposition: an ANOVA of the series on seasonal and trend factors reports the share of total variation explained by each component (the shares sum to one).
- [Smoothers](https://github.com/config-i1/greybox/wiki/Smoothers) — `lowess()`, `supsmu()` (R and Python). Stable. Non-parametric scatterplot smoothers (Cleveland LOWESS, Friedman SuperSmoother).

## Demand analysis

- [AID](https://github.com/config-i1/greybox/wiki/AID) — `aid()`, `aidCat()` (R) / `aid()`, `aid_cat()` (Python). Stable. Automatic identification of demand type with stockout / new / obsolete detection.

## Evaluation and measures

- [measures](https://github.com/config-i1/greybox/wiki/measures) — `measures()`, `ME()`/`MAE()`/`MSE()`/… (R) / `measures()`, `me()`/`mae()`/`mse()`/… (Python). Stable. Point, interval, quantile and half-moment error measures.
- [rolling_origin](https://github.com/config-i1/greybox/wiki/rolling_origin) — `ro()` / `rolling_origin()`. Stable. Rolling-origin cross-validation.
- [RMCB](https://github.com/config-i1/greybox/wiki/RMCB) — `rmcb()` (R and Python). Stable. Regression-based Nemenyi/MCB test for comparing forecasting methods: ranks the data, fits a dummy regression, and groups statistically indistinguishable methods (`distribution="tukey"` reproduces the Nemenyi test).
- [association](https://github.com/config-i1/greybox/wiki/association) — `association()`, `pcor()`, `mcor()`, `determination()`. Stable. Measures of association.
- [diagnostics](https://github.com/config-i1/greybox/wiki/diagnostics) — `outlierdummy()` / `outlier_dummy()`. Stable. Outlier detection and dummy construction.

## Feature engineering

- [manipulations](https://github.com/config-i1/greybox/wiki/manipulations) — `xregExpander()`, `xregTransformer()`, `xregMultiplier()`, `temporalDummy()` / `xreg_expander()`, `xreg_transformer()`, `xreg_multiplier()`, `temporal_dummy()`. Stable. Variable transformation utilities.

## R-only (Python port pending)

- `sm()` (scale model), `coefbootstrap()`, `cramer()`, `graphmaker()`, `spread()`, `tableplot()`, `detectdst()`, `detectleap()`, `polyprod()`, `dsrboot()` — see the parity matrix on [Home](https://github.com/config-i1/greybox/wiki/Home).