Normal Curves: Sexy Science, Serious Statistics

Bonus: Pheromones with commentary

Jan 26, 2026
They revisit the sweaty T-shirt study that spawned pheromone parties and DNA-based dating fads. They unpack HLA/MHC genes and how scent was used to suggest mate choice. Statistical curiosities pop up: bar chart misuse, correlated observations, within-person designs, and post-hoc subgrouping. They trace failed replications, lost data, and a meta-analysis that questions the original claims.
Ask episode
AI Snips
Chapters
Books
Transcript
Episode notes
ADVICE

Don't Use Bar Charts For Numerical Data

  • Always report numeric summaries and avoid using bar charts for numerical data because they hide distribution and force readers to guess.
  • Provide means, sample sizes, and variability so others can replicate and assess results.
INSIGHT

Why Correlated Observations Matter

  • Correlated observations arise when multiple measurements originate from the same unit, e.g., multiple shirt ratings per woman.
  • Ignoring these dependencies inflates apparent sample size and risks misleading significance.
INSIGHT

Three Sources Of Dependency In The Study

  • The sweaty t-shirt data had three dependency sources: multiple shirts per woman, paired (within-woman) comparisons, and men rated by multiple women.
  • Wedekind averaged replicates and paired correctly but failed to account for correlations by man, risking one man's ratings dominating results.
Get the Snipd Podcast app to discover more snips from this episode
Get the app