Data Analysis Guide

Output Components

1. Method Justification Paragraph (~80 words)

• Why this method is appropriate for this model and data type
• Cite the canonical source: Hair et al. (2019) for PLS-SEM; Anderson & Gerbing (1988) for CB-SEM; Hayes (2018) for PROCESS
• Address sample size fit: "PLS-SEM is recommended for complex models with smaller samples (Hair et al., 2019); with N=[n], this study meets the minimum 10-times rule"
• If CB-SEM: justify why multivariate normality assumption is met or addressed

2. Analysis Sequence (numbered steps)

For PLS-SEM:

1. Assess measurement model: Cronbach's Alpha, CR, AVE, HTMT
2. Assess discriminant validity: Fornell-Larcker criterion / HTMT ratio
3. Run structural model: bootstrapping (5000 iterations), path coefficients, t-values, p-values, R²
4. Assess model fit: SRMR < 0.08
5. Conduct blindfolding for predictive relevance (Q² > 0)

For CB-SEM:

1. Confirmatory Factor Analysis (CFA): fit indices (CFI, TLI, RMSEA, SRMR)
2. Convergent validity: AVE > 0.50, CR > 0.70
3. Discriminant validity: HTMT < 0.85
4. Structural model: path coefficients, significance, R²
5. Model comparison if alternative models tested

For Multiple Regression (OLS):

1. Assumption checks: normality, homoscedasticity, independence (Durbin-Watson)
2. Multicollinearity: VIF < 10, Tolerance > 0.10
3. Hierarchical regression: enter controls in Block 1, IVs in Block 2
4. Report: β, t, p, ΔR², F-change

For Mediation (Hayes PROCESS):

1. Direct effect (path c')
2. Indirect effect via bootstrapping: 5000 iterations, 95% CI
3. Report: a × b indirect effect, LLCI, ULCI (significant if CI excludes zero)
4. Model number: PROCESS Model 4 (simple mediation), Model 6 (serial), Model 7 (moderated mediation)

3. Fit Index / Threshold Table

4. Common Method Bias (CMB) Specification

• Test: Harman's single-factor exploratory factor analysis
• Threshold: single factor must not account for > 50% of total variance
• If Harman's > 50%: recommend marker variable technique or post-hoc statistical remedies
• Report in Results: "Harman's single-factor test revealed that one factor accounted for [X]% of variance, below the 50% threshold, indicating CMB is not a major concern"

5. Sample Size Adequacy Check

• PLS-SEM minimum: 10 × maximum number of paths pointing to any construct
• CB-SEM minimum: 200 (complex models) or 5–10 observations per indicator
• OLS minimum: 20 observations per predictor variable
• Report verdict: "The sample of N=[n] [meets / exceeds / marginally meets] the minimum requirement of [formula result]"