Skip to content
openentropy logo openentropy

Statistics Analysis

Statistics analysis provides classical tests for distributional fit and dependence.

Implemented in openentropy_core::statistics.

What it is: A classical hypothesis-testing toolkit (fit, dependence, and group inference).

Use it for: Formal p-value-driven evidence beyond heuristic forensic checks.

Input shape: One byte stream for core tests; multiple groups/streams for ANOVA/Kruskal/Levene.

Use this when

Section titled “Use this when”
  • You want classical hypothesis-test style evidence (p-values) for fit/dependence.
  • You need serial-correlation tests beyond forensic heuristics.
  • You are comparing groups/sessions with ANOVA/Kruskal/Levene workflows.

Single-stream methods

Section titled “Single-stream methods”
  • cramer_von_mises: uniformity goodness-of-fit test
  • ljung_box / _default: multi-lag autocorrelation significance test
  • gap_test / _default: interval gap structure against expected random gaps
  • statistics_analysis: one-call orchestrator for the single-stream statistics set

Group-level methods

Section titled “Group-level methods”
  • anova: parametric group mean test
  • kruskal_wallis: non-parametric group rank test
  • levene_test: equal-variance test across groups
  • power_analysis / _default: approximate power and required sample sizing
  • bonferroni_correction, holm_bonferroni_correction: family-wise correction

CLI

Section titled “CLI”
Terminal window
openentropy analyze --statistics
openentropy analyze --profile deep

Python SDK

Section titled “Python SDK”
from openentropy import statistics_analysis, ljung_box, anova, holm_bonferroni_correction


stats = statistics_analysis(data)
lb = ljung_box(data)
a = anova([group_a, group_b, group_c])
holm = holm_bonferroni_correction([0.001, 0.02, 0.2], 0.05)
Section titled “Related”