Sloppiness Makes a Good Story

February 23, 2012

I am usually suspicious when a journalist writes about statistics and studies conducted by social scientists.  Here’s a good example of why.

The flaws in this study can be quickly pointed out.  First, sleep-deprived is defined as having less sleep than other people.  This is a horrible definition, as some people can naturally function on less sleep than others and may be inclined to choose certain professions.  There are other possible reasons as to why the people in certain professions get less sleep but may be better-rested.

Second, the differences in time sleeping are so slight in percentage terms that claiming there is a significant statistical difference between any of the “top ten sleep-deprived jobs” would require an absurdly large sample size from the population at large.  27,157 people were sampled out of 300,000,000.  This does not strike me as being enough to really allow the standard error to be large enough so that these differences can be attributed to something other than chance.

Third, the jobs with most sleep (“well-rested”) also tend to do the most physical labor, so their bodies would need more sleep to be just as rested as other people.  This means that sleeping 23 minutes more than home health aides may not indicate which is more sleep-deprived.  I know from personal experience that when I am highly active during the day, I need at least seven hours of sleep, as opposed to the days when I do little or no activity and can get by on six hours of sleep and feel just as rested.

But maybe the erroneous conclusions about sleep-deprivation can be chalked up completely to the journalist who hasn’t a clue about how statistics work.  The headline is unwarranted by the data.  The amount of rest people have is not (on the whole) caused by the amount of sleep they get.  Working a more physical job does not mean one gets more rested–maybe it means less rested.

Overall, this bit of journalism is sloppy and a nice story to generate debate in the comments section.  Typical.