Rather than trying to draw policy conclusions from an epidemiological study like this, wouldn't it be more accurate to give drivers measured amounts of THC, put them in a driving simulator, and measure their performance?