Skip to main content

Dataset Comparison

Last updated: Apr 17, 2026, 6:46 AM EDT

This page provides a comprehensive side-by-side comparison of all four primary result groups. Each group represents an independent dataset with progressively less restrictive quality criteria. Researchers can use this comparison to assess how data cleaning decisions affect statistical conclusions.

Sample Overview

#Result GroupNDefinition
1Conservative Clean89Prolific APPROVED + all quality checks (IRI, duration >= 540s, reCAPTCHA, straightlining, auth)
2Flexible Clean140Prolific APPROVED + basic quality (all 3 IRIs + duration >= 480s)
3Prolific Accepted261All deduplicated V2 rows with Prolific APPROVED status
4All V2 Finished410Finished + duration >= 120s (extreme speeders excluded)

Core Metrics Comparison

All descriptive statistics, reliability coefficients, and correlations are shown for each result group. The Δ column shows the difference from Conservative Clean (the strictest dataset). Deltas > 0.05 are highlighted in amber.

MetricConservative CleanFlexible CleanProlific AcceptedAll V2 FinishedΔ FlexibleΔ ProlificΔ All
Barrier Grand Mean2.83542.81352.79442.7591-0.0219-0.0410-0.0763
Barrier SD0.62520.71150.70920.7658+0.0863+0.0840+0.1406
Readiness Grand Mean3.05203.08623.12603.2284+0.0342+0.0740+0.1764
Readiness SD0.56430.65730.67010.7194+0.0930+0.1058+0.1551
Maturity Grand Mean3.05263.06563.15303.2593+0.0130+0.1004+0.2067
Maturity SD0.69880.80640.80720.8074+0.1076+0.1084+0.1086
B-R Correlation-0.4265-0.4485-0.3457-0.3042-0.0220+0.0808+0.1223
B-M Correlation-0.1783-0.3141-0.2815-0.3189-0.1358-0.1032-0.1406
R-M Correlation0.57830.70650.72080.7235+0.1282+0.1425+0.1452
Alpha Barriers0.85350.87570.87640.8997+0.0222+0.0229+0.0462
Alpha Readiness0.86770.91710.91830.9317+0.0494+0.0506+0.0640
Alpha Maturity0.82910.88710.88990.8909+0.0580+0.0608+0.0618

Survey Demographics Comparison (Qualtrics)

Role distribution and organization size breakdown for each result group, based on self-reported survey responses (Q1, Q4). These are organizational demographics from the TABS instrument. Prolific platform demographics (age, sex, ethnicity, plus prescreener fields like industry, company size, and occupation) are collected separately and can be used to cross-validate these survey responses via Prolific Participant ID.

Tech vs Non-Tech Composition

Result GroupNTechnicalNon-TechnicalOther% Tech
Conservative Clean891970021.3%
Flexible Clean14034106024.3%
Prolific Accepted26161200023.4%
All V2 Finished410108302026.3%

Organization Size Distribution

Result Group<100100-499500-9991000-49995000-999910000+
Conservative Clean1131822611
Flexible Clean19441834817
Prolific Accepted457035532137
All V2 Finished5911962813356

Filter Bias Analysis

This analysis tests whether stricter quality filters disproportionately exclude certain demographics. A Chi-square test for independence is computed across the four result groups for role, organization size, and profit model.

Demographic CategoryChi-Square (χ²)dfp-valueInterpretation
Role (Tech vs Non-Tech)1.3930.7072No significant difference (demographics stable)
Organization Size8.37150.9079No significant difference (demographics stable)
Profit Model3.3860.7605No significant difference (demographics stable)

Profit Model Distribution

Result GroupFor-ProfitNon-ProfitGovernment/Public Sector
Conservative Clean631511
Flexible Clean952916
Prolific Accepted1924623
All V2 Finished3057035

Effect Size Comparison

Cohen’s d effect sizes for key group comparisons across all four result groups. This shows how effect sizes shift as the sample becomes less restrictive.

Tech vs Non-Tech (Cohen’s d)

ConstructConservative CleanFlexible CleanProlific AcceptedAll V2 Finished
barriers-0.093-0.264-0.045-0.132
readiness0.4380.5780.5550.455
maturity0.2630.3840.3630.321

Large vs Small/Medium Org (Cohen’s d)

ConstructConservative CleanFlexible CleanProlific AcceptedAll V2 Finished
barriers0.5150.3710.1570.162
readiness-0.065-0.1660.1000.069

Interpretation Guide

Metrics that remain stable across all four groups suggest robust findings that are not sensitive to data cleaning decisions. Metrics that show large deltas (highlighted in amber) between Conservative Clean and less restrictive groups warrant further investigation, as the finding may depend on sample composition.

As a rule of thumb: if a Cohen’s d shifts by more than 0.1 between Conservative Clean and All V2 Finished, the effect size may be inflated or attenuated by lower-quality responses. Similarly, if Cronbach’s α drops below 0.70 in larger samples, it may indicate that less-engaged respondents are adding noise to the scale.

Related