regression-diagnostics 0.2.0

Statistical diagnostics for OLS regression in Rust: VIF, condition number, adjusted R2, F/AIC/BIC, residual tests (Durbin-Watson, Breusch-Pagan, White, Jarque-Bera), influence measures (leverage, Cook's distance, DFFITS), QQ-plot data, and an R/statsmodels-style summary().
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25

//! Multicollinearity diagnostics: [`vif`] and [`condition_number`].
//!
//! These answer *different* questions and can disagree, so the crate exposes
//! both rather than treating one as redundant:
//!
//! * **VIF** is *per-predictor*: how inflated is coefficient `j`'s variance
//!   because predictor `j` is linearly explained by the others.
//! * **Condition number** is a *whole-design* scalar: how ill-conditioned the
//!   design matrix is overall. A design can have a large condition number driven
//!   by a combination of predictors while no single VIF is dramatic, and vice
//!   versa.
//!
//! ## Threshold guidance (convention, not mathematical fact)
//!
//! The following rules of thumb are widely cited but vary by source; treat them
//! as convention:
//!
//! * VIF **> 5** is often flagged as concerning, **> 10** as serious.
//! * A condition number **> 30** is a common concern threshold.

mod condition_number;
mod vif;

pub use condition_number::condition_number;
pub use vif::vif;