Hypothesis testing
H1- Maslow spread predicts Developmental Affordance (DA)
H1: There is a positive association between Maslow entropy (H) representing the spread of needs across L1–L6 and the Developmental Affordance (DA) index at the image level.
H0:There is no association between Maslow entropy (H) and the Developmental Affordance (DA) index at the image level.
Question: Do images that display a more balanced spread of human needs also tend to contain a richer variety of opportunities for human development?
Data
Data subset: Images containing at least one Maslow category and at least one DA category
Variables:
-
- Predictor(s):Maslow entropy (H) over six levels (L1–L6), per image.
- Outcome(s):DA index in [0,1], per image from 10 DA groups.
Controls: Category fixed effects via meta-analysis; within-image multiplicity weighting; inverse category frequency weighting.
Normalisation & Weighting: See Weighting and Normalisation Methods reference page.
N: 6,940
Methods
Test type:
-
Spearman rank correlation (ρ) and Kendall’s τ.
-
Weighted analogue of Spearman correlation.
-
Per-category correlations pooled with DerSimonian–Laird random-effects meta-analysis.
-
LOCO (leave-one-category-out) sensitivity analysis.
Rationale: Non-parametric rank correlations are robust to monotonic non-linear relationships and outliers. The weighted version incorporates design weights while preserving non-parametric ranking. Random-effects meta-analysis accounts for variation across search categories. LOCO checks for single-category influence.
Implementation: [Identical specifications used for binary and weighted datasets. Bootstrap 95% confidence intervals for ρ; weighted p-values from permutation under fixed weights.
Results — Weighted
Global Correlation
| Metric | Value | 95% CI | p (two-sided) | p (one-sided, H₁) |
|---|---|---|---|---|
| Spearman ρ | 0.1582 | [0.1356, 0.1804] | 2.00 × 10⁻⁴ | 1.00 × 10⁻⁴ |
Meta-analysis
| Model | Random-effects r | 95% CI | τ² |
|---|---|---|---|
| Weighted | 0.1259 | [0.0877, 0.1637] | 0.00484 |
Interpretation: A small-to-moderate positive association between H and DA index, significant and consistent across categories even after balancing.
Results — Binary
Global correlation
| Metric | Value | 95% CI | Kendall τ | p (two-sided) | p (one-sided, H₁) |
|---|---|---|---|---|---|
| Spearman ρ | 0.2191 | [0.1976, 0.2404] | 0.1681 | 9.99 × 10⁻⁸¹ | 4.99 × 10⁻⁸¹ |
Meta-analysis
| Model | Random-effects r | 95% CI | τ² |
|---|---|---|---|
| Raw | 0.1805 | [0.1357, 0.2246] | 0.00824 |
Interpretation: Slightly stronger positive association than the weighted analysis, still consistent across categories.
Combined Interpretation
Both binary and weighted models show statistically significant positive correlations between H and DA index:
-
-
- ρ_raw ≈ 0.2191
- ρ_weighted ≈ 0.1582
-
The attenuation in the weighted model reflects the balancing of large categories and dense images. Low τ² values show minimal heterogeneity. LOCO analysis confirms no single category drives the results.
Conclusion
H₀ rejected in favour of H₁.
Images with a more balanced spread of Maslow needs also tend to contain a richer variety of developmental affordances. This supports the idea that when needs are more evenly represented, opportunities for fulfilment also surface in the visual record a finding that aligns with the broader research focus on instinctive expression and suppression.
Notes & Next Steps
-
Correlation does not imply causation; directionality remains unknown.
-
Consider continuous DA richness (beyond index presence) in follow-up tests.
-
Potential category-content bias should be explored further.