savvi

savvi is a package for Safe Anytime Valid Inference (SAVI). Also, it’s a savvy pun.

The goal of this package is to provide methods to calculate confidence intervals and p-values that are valid at all sample sizes.

More formally, given a significance level \(\alpha \in [0, 1]\), a \((1 - \alpha)\) confidence sequence for parameters \(\theta \in \mathbb{R}^p\) is a sequence of intervals \(C_n\) such that

\[P(\forall n \geq 1 : \theta \notin C_n) \leq \alpha.\]

\(p_n\) is an anytime-valid p-value for that process if

\[P(\forall n \geq 1 : p_n \leq \alpha) \leq \alpha.\]

See Ramdas et al. (2023).

Install

Stable version:

pip install savvi

Development version:

pip install git+https://github.com/assuncaolfi/savvi

For development, use pdm.

Get started

See the example gallery.

References

Ramdas, Aaditya, Peter Grünwald, Vladimir Vovk, and Glenn Shafer. 2023. “Game-Theoretic Statistics and Safe Anytime-Valid Inference.” https://arxiv.org/abs/2210.01948.