Discovery

Find patterns the framework didn't hypothesize.

The matrix shows Spearman rank correlation (ρ) between all numeric and ordinal variables — more appropriate than Pearson for mixed ordinal scales. Variables use different scales: 0-5 sliders (Productive, QoL, Completion, Drive, Wellbeing, Intro/Extra) and 1-5 frequency (Overwhelm, Overcommit, Priority Difficulty, External Pressure, Replan). ρ ranges from −1 to +1: near +1 = move together, −1 = move opposite, 0 = no relationship. Bold values are statistically significant (p < 0.05). Gray values may be noise. Three categorical dimensions (Hours, Tools, ND Impact) are ordinal-encoded so they can participate in correlation analysis. Hover any cell to see each variable's scale type.

Correlation Matrix

Top Correlations

The 10 strongest pairwise relationships, ranked by |ρ|. Significant correlations (p < 0.05) are solid; non-significant are faded — they may be noise given the sample size. Shows which problems cluster together.

Notable Findings

Every significant correlation with |ρ| ≥ 0.4 — each backed by a distribution heatmap and qualitative evidence from respondents living at the extremes. These relationships emerged from the data, not from prior hypotheses.

Role Profiles

How do people in different roles score on every metric? Each row is a role, each column a metric. Respondents appear in every role they hold (avg 4.2 roles/person). Darker = higher score. Read across to see a role's pain fingerprint; read down to see which roles struggle most with each dimension.

Role Co-Occurrence

Which roles travel together? Brighter = more often paired. People wearing highly co-occurring role combos experience compound context-switching costs.

Worst-Outcome Role Combinations

For each pair of co-occurring roles, mean outcome metrics compared to the population baseline. Sorted by highest overwhelm. Only pairs with N ≥ 10.

Categorical Breakdowns

How do key metrics differ across categorical segments? Pick a grouping dimension and a metric to explore. Box plots show median (line), IQR (box), and range (whiskers) — revealing distributional differences that medians alone miss.