6 ms·
Be very wary of headlines like this. They draw the line at 6 hours, but that includes everyone, even people that only get 3 hours. What we want is the raw curve
by 3pt14159 14y ago
Be very wary of headlines like this. They draw the line at 6 hours, but that includes everyone, even people that only get 3 hours. What we want is the raw curves. If you get exactly 5 hours and 55 minutes a night, you might only be 1.5 times as likely as someone who gets 9 hours a night, while the people getting 4 hours a night are 15 times more likely.
- dredmorbius 14y agoWhat you want to see is the study, how the analysis variables were constructed, and what models were tested. If a simple binary "subject slept 6 hours or fewer" variable is constructed, and there's no other discriminant, then if there is a significant positive correlation, it will show. Testing multiple conditions, exclusively (3 or fewer, 4 or fewer, 5 ..., etc.) would get around this. By way of comparison: I've used the Harris-Benedict basal metabolic rate formula to estimate caloric expenditure by individuals. The question came up: how well does this scale to extreme weights. It turns out that the original study is online as a PDF, including full stats on all the study subjects. My surprises were that 1) the formula is a lot older than I'd imagined (dating to 1919), involved a relatively small study size (a few score participants), and the weight range did not include high-end extremes unfortunately rather too common in current populations. I've taken to treating the values generated for extremely overweight individuals with a bit more suspicion as a result. Study context matters.