Hierarchical Regression — Sequential blocks with ΔR² + F-change
VerifiedAdvanced & specialized
Run this test straight away on a free built-in teaching dataset — no data of your own needed — or bring your own. Either opens the guided workspace: variable setup, assumption diagnostics, results with effect sizes and confidence intervals, figures, and APA-ready reporting.
Loading teaching datasets…
Or use your own dataset
Loading your datasets…
Enters predictors in researcher-specified ORDERED BLOCKS and tests the incremental variance (ΔR²) added by each block BEYOND the preceding blocks.
Reports per-block R², adjusted R², ΔR², F-change with df (df₁ = block size, df₂ = residual df), p-change, plus the final-block coefficient table with full diagnostics. Block order is THEORETICAL, not algorithmic — this is NOT stepwise regression. The standard tool in psychology / education research for testing whether a focal predictor adds explanatory power beyond established controls.
Worked example
Does personality add predictive power beyond study time?
Block 1 enters study time; Block 2 adds conscientiousness. Hierarchical regression tests whether the second block improves prediction of exam score.
Study time alone gave R² = .18. Adding conscientiousness raised it to R² = .37 — a significant gain, ΔR² = .20, F-change(1, 87) = 27.1, p < .001.
Adding conscientiousness significantly improved prediction of exam score beyond study time, ΔR² = .20, F(1, 87) = 27.1, p < .001.
Try it yourself: Load this ready-made sample and follow the run above.
When to use it
- Demographic / clinical controls → focal predictor → moderatorExam score ~ Block 1: age + prior GPA (R² = .42); Block 2: hours studied (R² = .47, ΔR² = .05, F-change = 14.3, p < .001) — study time matters even after ability proxies.
- Incremental validity of new psychometric measureJob performance ~ Block 1: cognitive ability + conscientiousness (R² = .31); Block 2: new emotional intelligence scale (R² = .35, ΔR² = .04, p = .002) — EI adds incremental validit
When NOT to — use instead
- Single block of predictorsJust use linear regression — no blocks to compare. → Linear Regression \u2014 OLS (continuous y + predictors)
- Stepwise / data-driven model selectionHierarchical requires THEORETICAL block ordering.
- Binary outcomeUse hierarchical logistic regression (same block-comparison logic on the LL scale). → Binomial Logistic Regression \u2014 odds ratios + ROC AUC + classification
- Clustered / nested dataUse hierarchical linear models (HLM) or mixed-effects to handle the nesting structure. → Linear Mixed Effects (LMM) \u2014 clustered / repeated data (ICC, BLUPs)
Hypotheses
Parameter tested: ΔR² (incremental) per block; vector of population coefficients β
Assumptions (and what to do if they fail)
Check: Shapiro-Wilk p-value on final-block residuals.
If violated: Non-normal residuals affect small-n CIs / p-values; CLT protects for n ≥ 100.
Check: Breusch-Pagan p-value.
If violated: Heteroscedasticity inflates SEs — F-change tests and per-coefficient inference become unreliable.
Check: Durbin-Watson statistic in [1.5, 2.5] is OK.
If violated: Autocorrelated residuals bias SEs and the F-change test.
Check: Max Cook's D reported.
If violated: Influential cases can flip the sign of ΔR² conclusions block-by-block.
Check: Max VIF across predictors.
If violated: High VIF inflates SEs — significant ΔR² with no significant β shows shared variance among predictors in the same block.
Ready to run a Hierarchical Regression — Sequential blocks with ΔR² + F-change on your own data?
Guided setup, automatic assumption checks, effect sizes, figures and an APA write-up.
Run this test →