rcrpy

Robust Chauvenet Rejection (RCR) — advanced, but easy-to-use, statistical outlier rejection that stays accurate even on samples that are more than 85% contaminated. rcrpy is the pure-Python reimplementation of the RCR algorithm from Maples et al. 2018.

Installation

pip install rcrpy

User guide

• Single-value RCR
  • The four rejection techniques
  • Unweighted, iterative rejection
  • Bulk rejection (faster on large N)
  • Weighted rejection
  • Results
  • Performance
• Functional-form RCR
  • Minimal example: linear fit
  • Other model shapes
  • With error bars
  • With weights
  • Multi-dimensional independent variable (ND)
  • Performance
• Advanced functional-form features
  • Priors
  • Pivot points
  • Parameter uncertainties
  • MEDIAN / MODE mu_tech for parametric
• Migrating from legacy rcr to rcrpy
  • Imports
  • Constants and enums
  • Methods (camelCase → snake_case)
  • Result fields (camelCase → snake_case)
  • FunctionalForm
  • Priors
  • NonParametric
  • What rcrpy does NOT include from legacy rcr
  • Performance

Reference

• API reference

Development

• Publishing rcrpy