Comment Re:Statistical Magic (Score 1) 328
"The data showed that 7.6% of crash-involved drivers tested positive for marijuana or THC, versus 6.1% of the control group. In raw terms, that would suggest that marijuana was associated with a 25% increased chance of crashing. But it's not that simple: the figures have to be adjusted for other factors possibly contributing to crash risk, including the driver's gender and age."
So their control group wasn't representative of the "young male" population.
Their "young male" correlation is also subject to a VERY large p-value (0.65), which is a problem in itself. The chart showing the "Adjusted Odds Ratios Between Drug Class Use and Crash Risk (Adjusted for Demographic Variables: Age, Gender And Race/Ethnicity)" is notable for some VERY high p-values.
The rest of the paper considers a p-value of 0.05 to be significant: a value of 0.65 falls into "we don't even believe this ourselves" territory.