Sandbox

Reproducibility

When studies are wrong: A coda
by George Johnson, New York Times, 7 March 2014

Submitted by Bill Peterson

More on Kansas

The following web site reversed the axes and its regression line is correct:

http://www.mischiefsoffaction.com/2014/04/are-democrats-pervs-some-problems-with.html

so we obtain the following graph

This "reverse causality" has porn viewing "causing" the Obama vote whereas the original has Obama vote "causing" porn viewing. Instead of the loaded term, "causing," this web site tongue-in-cheek says

Porn pageviews explain 16 percent of the variance in state level presidential vote shares. Each per capita pageview is associated with a two-tenths of a percentage point increase in a state's Democratic vote share, and this is statistically significant at the p≤.01 level.