Beware Summary Statistics

The short predecessor to 'Stalin and Summary Statistics.' Two Tumblr-sourced puzzles: (1) IQ↔crime at r=−0.19 (3.6% variance) sounds trivial, yet is consistent with all crime being low-IQ — because r mixes p(low-IQ|criminal) with p(criminal|low-IQ), and the latter is tiny since few people commit crimes; (2) IQ↔income at r≈0.4 (16% variance) sounds minor yet the decile means span $40k–$160k (huge within-decile variance). The moral: beware summary statistics — the same data looks important or trivial depending on which statistic you cite; when in doubt, demand the scatter plot.

Quality 67 · Strong. Strong, low. A clear, genuinely useful statistical-literacy point — the p(A|B)≠p(B|A) insight on the crime example is sharp. Held low-Strong because it's brief (612 words) and was later expanded into the fuller, better SSC-411 (72); this is the seed.

Claude’s paradigm shift 48 · Moderate. Moderate (48). The 'beware summary statistics / demand the scatter plot' framing was a useful public-facing point at publication, building on well-known statistics.

Real-world impact 2 · Minor. Minor (2). Improves readers' statistical literacy; the 'demand the scatter plot' point circulates within the discourse; no material reach.