E-Cig Study: Much Smoke, Little Light

A clean statistical critique of a study finding e-cigarette users no more likely to quit smoking. Scott's explanation is length-biased / survivorship sampling, taught via a 'Hypothetical World' of Short Smokers (smoke 1 year) and Long Smokers (50 years): sampling at a tobacco shop catches ~50x more Long Smokers, so you'd wrongly conclude Long Smokers are 50x more common. Applied to e-cigs: successful users switch fully within a month (rarely caught in a tobacco-smoker sample), while unsuccessful users 'futz' for years (heavily over-represented) — so a cross-sectional sample of smokers over-represents 'the group even e-cigs can't help,' explaining the null/negative result. A speculative but clear demonstration of a subtle sampling bias.

Quality 66 · Strong. Strong, low (66). A clean, instructive demonstration of length-biased sampling applied to a real study, with a memorable teaching device (Short vs Long Smokers). Held to low-Strong by its brevity, single point, and explicitly speculative framing.

Claude’s paradigm shift 48 · Moderate. Moderate (48). A fresh, clarifying application of length-biased sampling to the e-cig finding; the bias itself is a known concept.

Real-world impact 2 · Minor. Minor (2). A within-blog statistics-teaching critique; no policy or practice effect.