Work

The Silence Experiment: Phase 0 Statistical Power Analysis and Study Design Table

Year: 2026
Type: article
Abstract: Two statistical companion documents to The Silence Experiment, deposited together because they answer the same question from two directions: what the protocol can detect, and how each pre-registered claim maps onto a sampling plan, an analysis plan, and an interpretation fixed before any data exist. The power analysis computes achievable power for both planned contrasts across the design space, using the non-central t distribution with degrees of freedom set by participant count, treating the participant as the unit of inference. The central Phase 0 finding is that participant count, not session count, is the binding constraint, and that the person-to-person heterogeneity of the deep-versus-shallow effect must be estimated from pilot data before the stopping rule can be locked. Minimum detectable effects are reported so that a null can be stated as a quantified upper bound rather than as an absence, and the analysis states plainly that every such bound is conditional on the pre-registered positive control passing, since a null from a chain that could not have detected anything carries no information about effect size. A section on the redefined Phase 1 reports what a reduction in per-session noise buys. The result that matters is that the second contrast, which no feasible Phase 0 configuration brings above the stopping rule, clears it under a modest noise reduction on the same platform with the same participants. The first contrast gains only where it is weakest and saturates elsewhere, because between-person heterogeneity is a floor that no reduction in measurement noise touches. Both readings are stated. The section is explicit that the noise reduction is a sensitivity analysis over an assumed improvement rather than a prediction of one, that a countervailing mechanism exists, and that the governing quantity is measured at commissioning rather than assumed. The study design table follows the Registered Reports template: research question, hypothesis, sampling plan, analysis plan, and interpretation given different outcomes, for each pre-registered hypothesis including those belonging to Phase 1, plus the auxiliary assumptions covering the frozen EEG classifier, environmental isolation, and the sensitivity control. All contrast weights are stated explicitly with their sums. The table is design-level and host-independent: facility, ethics reference, enrolled participant count and final session count belong in the manuscript body once a host institution is secured. Neither document is peer reviewed and no data have been collected. Every figure is computed rather than cited, and the two documents are locked to the same design points and the same values as the corresponding table in the protocol paper so that they cannot drift apart. Changes in this version. Secondary test D is stated with corrected weights that sum to zero, filed as a pre-data amendment with its rationale. Rows are added for the two new Phase 1 hypotheses. All references to the withdrawn transmon phase are repointed to the redefined Phase 1. The session model reflects chained conditions. The power analysis gains a section on what Phase 1 buys, a note that the assumed per-session noise variance is imported from a general literature rather than derived from this instrument, and a recommendation that it be measured at commissioning instead. No repository identifiers appear in either document.
Language: English
Open Access status: green
OpenAlex ID: W7167787715
FWCI:
Cites:
Cited by:
Related to: 10