spurious correlations

Abstraction: Database of thousands of statistically significant but causally meaningless correlations

Key points:

• Tyler Vigen's site compares 25,237 variables against each other — 636,906,169 correlation calculations — to surface random high correlations (data dredging since 2014).
• Any sufficiently large dataset will yield strong correlations by chance; starting with data rather than a hypothesis is a core methodological danger.
• Sequential years are not independent observations, inflating p-values; naive degrees-of-freedom counting misrepresents the true probability of random agreement.
• Confounding variables (e.g., global pandemics) can make two unrelated variables appear connected; outliers can substantially inflate single correlations.
• Visual tricks compound the effect: truncated Y-axes and line graphs make weak or coincidental relationships look dramatic; clicking any chart reveals a version starting at zero.

Connections: Tyler Vigen · Spurious Correlation · Data Dredging · Statistical Fallacies

Source: https://www.tylervigen.com/spurious-correlations