LearnMinds
Specialized Hub
Advanced Skill Acquisition

Stat Plan
Mind

The Blueprint of Scientific Truth

The Vision

Case Narrative

The Vision

A Statistical Analysis Plan (SAP) is the ultimate binding contract between a researcher, their data, and the scientific method; it ensures that findings emerge from pre-specified logic rather than post-hoc statistical creativity. The core principle of pre-specification—'Decide Before You See'—forces investigators to make all analytical decisions prior to viewing outcome data, protecting against both conscious and unconscious biases that systematically erode research credibility. In modern empirical science, clarity in analysis is the only reliable safeguard against the reproducibility crisis, where unchecked analytical flexibility can cause 50–90% of published findings to be false, exaggerated, or irreproducible. A rigorously designed SAP transforms statistics from a tool of persuasion into a tool of truth.
First Principle

Systemic Failure Audit

Systemic Failure Audit

Systemic Failure Audit

Status

Active Critical Scanning

HIGH RISK
90%

50-90% of published research findings may be false or exaggerated due to high levels of analytical flexibility and undisclosed researcher degrees of freedom.

HIGH RISK
5%

Researcher degrees of freedom—choosing between multiple models, covariates, or outcomes—can inflate false positive rates from a nominal 5% to over 60% when multiple approaches are tested without correction.

HIGH RISK
36%

Only 36% of 100 landmark psychology studies could be successfully replicated in major open science initiatives, exposing the fragility of post-hoc analytic practices.

HIGH RISK
31%

31% of clinical trials are found to switch primary outcomes after seeing non-significant results, a practice that invalidates statistical inference and misleads clinicians.

CATASTROPHIC
!!

Scientific Retraction: Discovery of post-hoc 'outcome switching' or 'p-hacking' is a primary driver of high-impact journal retractions and institutional investigations.

CATASTROPHIC
!!

Statistical Dilution: Hunting for 'significant' subgroups when the overall effect is null creates false clinical hope and fuels wasteful downstream research.

CATASTROPHIC
!!

Regulatory Rejection: Regulatory bodies like the FDA or EMA will reject trial data if the final analysis deviates from a pre-registered SAP without transparent, rigorous justification.

The Deadly Sins

The Deadly Sins

Detection & Mitigation Protocol
1

Outcome Switching

"Changing the primary outcome to a secondary one because the primary failed to reach significance, thereby retrofitting success."

Elite Neutralization

Explicitly pre-register the one and only primary outcome before data collection begins.

2

Subgroup Fishing

"Hunting for specific groups (e.g., 'men over 50') that show significance when the overall trial failed."

Elite Neutralization

Define all subgroup analyses a priori and strictly limit their number to maintain statistical power and credibility.

3

Covariate Fishing

"Testing multiple demographic or clinical adjustments to find the 'right' model that produces p < 0.05."

Elite Neutralization

List exactly which covariates will be included in the final model before unblinding.

4

Missing Data Imputation Post-hoc

"Choosing an imputation method (like LOCF vs Multiple Imputation) based on which one improves the results."

Elite Neutralization

Declare the missing data handling strategy in the SAP before viewing dropout patterns.

5

Ignoring Assumptions

"Failing to plan for what happens if data is non-parametric or violates test assumptions."

Elite Neutralization

Define a decision framework for switching from parametric to non-parametric tests based on pre-specified normality criteria.

6

Multiple Comparison Error

"Testing dozens of endpoints without adjusting for the increased risk of false positives."

Elite Neutralization

Apply Bonferroni or False Discovery Rate (FDR) corrections to all secondary analyses.

7

Analysis After Unblinding

"Writing the analysis code after knowing which group is which, allowing for subtle 'interpretive drift'."

Elite Neutralization

Analysts should remain masked (Group A vs Group B) until the final analysis code is locked and executed.

Technical Standards

Protocol Intelligence

Technical Standards

locking protocol
  • Draft the SAP before data collection (Prospective) or at minimum before unblinding (Retrospective).
  • Register the SAP on Clinical Trials.gov or OSF to create a permanent, public timestamp.
  • Submit the SAP as an appendix to the IRB/Ethics Committee during the initial application.
  • Use strict version control with PI and Statistician signatures if public registration is restricted.
decision frameworks
measurement temporal
Pre-specify the exact timing of follow-up assessments used for primary analysis.
parametric vs nonparametric
Define numeric criteria (e.g., Shapiro-Wilk p-value < 0.05) that will trigger a switch in test types.

Readiness Checklist

Mission Readiness Protocol

Readiness Checklist

0/6
Verified Units

Implementation Playbook

Implementation Playbook

1

phase 1 pre specification

Define the primary outcome, timepoint, and estimand before any data access. List all planned covariates with clinical justification. Specify handling of missing data (MI vs complete case).

2

phase 2 protocol lock

Upload SAP to Clinical Trials.gov or OSF. Obtain PI and Statistician signatures. Freeze analytic decisions before unblinding.

3

phase 3 analysis design

Predefine primary statistical model (e.g., linear mixed model, Cox regression). Define criteria for assumption checks (normality, proportional hazards). Specify adjustment for multiple comparisons.

4

phase 4 masked analysis

Analyst works with Group A vs Group B labels only. Code is finalized without seeing treatment assignments. Independent statistician reviews the analytic code.

5

phase 5 unblinding and execution

Unblind only after analysis code is locked. Run final analysis exactly as pre-specified. Document any deviations with justification.

6

phase 6 reporting

Report both pre-specified and exploratory analyses clearly labeled. Include sensitivity analyses for robustness. Publish SAP alongside manuscript as supplementary material.

7

phase 7 audit and reproducibility

Archive raw code, data dictionary, and version history. Enable independent replication where possible. Maintain transparent documentation for regulatory review.

Foundational Methodology

Protocol Intelligence

Foundational Methodology

the validity nexus
title
Pre-specification & Internal Validity
concept
A statistical analysis plan written AFTER seeing the data is not a plan—it is retrospective rationalization dressed as methodology.
protective function
Locking the SAP prevents 'Hypothesis Shifting' and 'Covariate Fishing,' where investigators iteratively test multiple adjustments until one yields a p-value < 0.05.
threats to neutralize
p hacking
Trying multiple analytical approaches until significance is found; with 20 different analyses, the false positive rate climbs to ~64%, making chance look like truth.
harking
Hypothesizing After Results are Known; presenting exploratory or accidental findings as if they were the primary predictions of the study.

Canonical Foundations

Canonical Foundations

Authority & Lineage Audit
REF 01
foundational texts

"Altman — Practical Statistics for Medical Research"

Verified Source
REF 02
foundational texts

"Gelman et al. — Bayesian Data Analysis"

Verified Source
REF 03
foundational texts

"Hernán & Robins — Causal Inference: What If"

Verified Source
REF 04
foundational texts

"Harrell — Regression Modeling Strategies"

Verified Source
REF 05
foundational texts

"Moher et al. — CONSORT 2010 & Extensions"

Verified Source
REF 06
foundational texts

"Ioannidis — Why Most Published Research Findings Are False"

Verified Source
REF 07
foundational texts

"Rubin — Multiple Imputation for Nonresponse in Surveys"

Verified Source
REF 08
foundational texts

"Little & Rubin — Statistical Analysis with Missing Data"

Verified Source
REF 09
foundational texts

"Cochrane Handbook for Systematic Reviews of Interventions"

Verified Source
REF 10
foundational texts

"Senn — Statistical Issues in Drug Development"

Verified Source

The Final Truth

The Final Truth

The statistical analysis plan is not paperwork; it is the contract you make with the scientific method to ensure your findings are discoveries, not inventions. When rigor is sacred, every p-value represents a step toward legitimate medical truth.

100%