## Saturday, March 3, 2012

 Why didn't they use a multiple bar chart?
Joanne Jacobs says "While federal guidelines recommend 9.25 hours of sleep each night for students, that could be too much for teens, concludes a new study, which correlated students’ hours of sleep to reading and math scores."
The study found the optimal sleep amount for 10-year-olds ranges between 9 and 9.5 hours, while for 18-year-olds it is slightly less than 7 hours. At ages 12 and 16, children need between 8.34 to 8.43 hours and 7.02 to 7.35 hours, respectively, the study found.
First, very few people are capable of determining how long they stay asleep because they have no way of knowing when they fell asleep; the study was based on numbers that were self-reported (by children!). "There is heaping of sleep hours in the data on 0s and 5s. The distributions are computed with a smoothing parameter of 0.85." If you can't measure sleep time empirically, your data becomes suspect. To be fair, the researchers acknowledge "The first issue concerns the non-experimental nature of our data. Because our data were not generated by a controlled experiment, we cannot rule out the possibility that we are measuring a correlation between sleep and test scores rather than a causal relation."

 While this is interesting, it's not too useful.
Second, there was no control for studying time. Were those students studying if they slept less? Maybe the 18ers who are getting 6-7 hours of sleep are the ones studying and learning for hours while those who are sleeping for 9.5 hours are working less hard. Maybe no one is studying. Maybe they all are. Maybe the school should start at 8:30 instead of 7:30 (with a one-hour bus ride before that).
"To determine the optimal amount of sleep, they compared standardized test scores in mathematics and reading with the self-reported number of hours students were typically sleeping each night. "
Without knowing WHY the students are sleeping more or less, this is just another useless education study, listed in the advance online publication section. The results were released and everyone jumps on it to further some educational fad or just to make copy deadlines. "The study's results raise the issue of whether students receiving too much sleep may see a reduction in academic achievement. While more research is needed, the authors conclude that this is possible." Actually, the researchers don't get anywhere near that definitive a conclusion.

Let's look at their aggregate data; maybe that will clarify things.

The hours of sleep range from 4 to 12, with σ = 1.45 and =8 so we have a pretty narrow window here but it's still worth exploring. Income isn't ... wait ... mean income is 67k, range is 1.4 million, σ=\$92,000 ! Wow, I sure hope they controlled for income, but a look through the report doesn't show any indication that they did.

I know where I'd start round two.