Skip to content
openentropy logo openentropy

Trial Analysis

This document defines the session-trial analysis implemented in openentropy-core::trials.

Scope

Section titled “Scope”

OpenEntropy uses the term “PEAR-style” as a historical reference to the Princeton Engineering Anomalies Research (PEAR) methodology. This project is not affiliated with or endorsed by PEAR.

Core Trial Model

Section titled “Core Trial Model”
  • Trials are fixed-length Bernoulli samples, default N = 200 bits per trial.
  • For each trial:
    • ones = number of set bits.
    • Z_trial = (ones - N/2) / sqrt(N/4).
  • For a run of k trials:
    • cumulative deviation = sum_i(ones_i - N/2).
    • terminal Z = cumulative_deviation / sqrt(k * N/4).
    • effect size = terminal_z / sqrt(k).

This matches the normal/binomial framing used in PEAR-era REG analyses for 200-sample trials.

Cross-Session Combination

Section titled “Cross-Session Combination”

When combining multiple sessions, OpenEntropy uses weighted Stouffer composition:

Z_combined = sum_i(w_i * Z_i) / sqrt(sum_i(w_i^2)), with w_i = sqrt(n_i).

Here n_i is the number of trials for session i. Zero-trial sessions are excluded from composition.

Calibration Gate

Section titled “Calibration Gate”

The optional record --calibrate gate checks per-source baseline suitability before recording:

  • |terminal_z| < 2.0
  • bit_bias < 0.005
  • shannon_entropy > 7.9 bits/byte
  • std_z in [0.85, 1.15]

These are practical baseline constraints for rejecting strongly biased or unstable sources before trial-driven experiments.

Primary Historical References

Section titled “Primary Historical References”

  1. Jahn, Dunne, Nelson, Dobyns, and Bradish (1997), Correlations of Random Binary Sequences with Pre-Stated Operator Intention: A Review of a 12-Year Program. https://www.pear-lab.com/pdfs/1997-correlations-random-binary-sequences-12-year-review.pdf

  2. Jahn, Dunne, and collaborators (2000), Mind/Machine Interaction Consortium: PortREG Replication Experiments. https://www.pear-lab.com/pdfs/2000-mmi-consortium-portreg-replication.pdf

  3. Dobyns, Dunne, and Nelson (2004), The MegaREG Experiment. https://www.pear-lab.com/pdfs/2004-megareg-replication-interpretation.pdf

  4. Jahn and Dunne (1991), Count Population Profiles in a Random Event Generator Experiment. https://www.pear-lab.com/pdfs/1991-count-population-profiles.pdf

PEAR publications index: https://www.pear-lab.com/publications