I know that when we run A/B tests on our website, and calculate confidence intervals, we often need thousands of conversions before we see results that we are convinced are statistically significant.
Arrghh. The quintessential anova experiments by cobb were done on 8 data points.
All groups are different. Let me repeat. All groups have different means. The magnitude of the difference (difference between means) is called the effect size. If the effect is large you need a trivial amounts of data. If the effect is small you need lots.
The only thing p is telling you is whether you can reliably tell the datasets apart. It's NOT telling you how far the datasets are apart, which is what you really want to know - does this change cause a large difference.
15 years ago a circadian rhythms expert told me that early risers are people whose natural rhythm is shorter than a day. Night owls are people whose natural rhythm is longer than a day. Since the average human cycle is 25 hours, more people tend to be night owls than early birds.
When I was a kid I could not get to sleep unless the room was totally dark. Even with heavy curtains and so on it was still common for me to spend several hours lying awake in the dark. In fact I wrecked my eyes by opening the door a crack and reading covertly in the dim hall light that was on until my parents went to bed, just to fill up the tine.
The one befit of this insomnia is that I kind of learned where my triggers are sufficiently well to induce sleep if I absolutely must be up early for some reason. But frankly, it's easier for me to pull an all-nighter, especially if I'm going to be on a plane or something where I can reasonably go to sleep as soon as I'm strapped in.
My solution is to drape an old shirt over my eyes. Works great. There are also miniature pillows that are sold for the same purpose, but they are not as comfortable as the shirt for me.
I would be interested to learn about the importance of fitness and nutrition in the lives of people in the study. Proper exercise and dieting leads to increased energy and can make someone an earlier riser and a night owl...
Well, I'm definitely an early riser (before 6AM on most days), and I get some degree of exercise on most days (12 miles of cyclo-commute on office days, longer rides on other days), and I'm generally good 'til about 9:30PM. After that, I definitely start to crash.
However...if I go to the climbing gym at, say, 6:30PM or 7PM, and stay for 90 minutes, I can make it a little further into the night without being tired.
I'd like to know if this stuff is congenital, or if it can be changed by altering your sleep cycle.
In other words, if someone who is naturally a night owl becomes an early-riser by force of practice, will they retain this "second wind" in the evenings?
You won't retain the second wind if my experience is anything to go by.
With my 2am-9:30am sleep cycle, I could stay alert until pretty much any time I liked (very useful for writing a PhD, etc). Now that I am forced to wake up at 7:30am, I find myself crashing at 10pm.
Same for me. 2am - 9am is fine for me. But when I go to bed at 10pm I don't want to get up until 8am. If I force myself to get up at 6am then I'll be pretty much spent by 8pm as far as any serious mental work goes.
It's easy to figure out what is happening here: early risers wake up with the alarm clock. This means they're probably running some sleep debt.
I think late risers listen to their body and get more sleep. This leads them to be more alert, because they're not running a deficit. They might need 1/2 hour more sleep, and get it (when needed).
Obviously it's beneficial, we can see much easier in the daytime which is a considerable plus when it comes to searching for food.
As a night person, I find it harder to explain why I feel more productive after the sun has gone down. There is some support now for the idea that this can be an actual medical condition in some cases (http://en.wikipedia.org/wiki/Delayed_sleep_phase_syndrome) affecting <1% of the population, though about ~15% of people are thought to be night owls with a general preference for the evening hours but not inability to sleep earlier if they choose.
Though it's not something I, um, lose sleep over, I do think that society inadvertently discriminates against people with an evening orientation.
I find it horrific that many American schoolchildren are expected to begin lessons at 8am - unless I am working really long days where I'm worn out physically as well as mentally (so I sleep early), I generally can't function at all before 10am. In Ireland school starts at 9am and my childhood memories are dominated by difficult wakening, rushed breakfasts and persistent lateness. I have major reservations about how I'm going to handle this when we have kids, which will be soon.
I agree that school starts excessively early. I have to wake up by 7am to get to school by 8:15am. School started this week, so I'm in a good position to observe this - during the summer I stayed alert until at least 2-3am. I woke up around 10am. Now that I'm getting up earlier, I feel tired around 9:30-10:00pm and have trouble staying awake. That's a loss of several hours for me. Although I can wake up in the morning fully rested, I'm not alert or focused until closer to lunch time, and definitely not until 9.
Effective artificial light hasn't been around long enough for any evolutionary adaptations to it. Normally, people were awake when it was light and slept when it was dark. When camping, I normally wake at the crack of dawn and go to sleep pretty early after sunset. At home, when not working (which means without a fixed schedule) I tend to sleep from 4 AM to noon. When working I can adapt fairly easily to any working shift, but without the external focus I always drift back to the 4 AM to noon. I have tried to maintain other schedules (as I have mentioned previously on HN posts on sleeping), but they never work out for very long.
As for the point of the article, sleeping the 4 AM to noon shift seems to be stable for me because I never seem to "crash" on that schedule as I sometimes do on others. I can stay up much longer than 4 AM, even completely miss a sleep period, and still be back to 4 AM to noon the next day; I cannot do that with other sleep schedules.
You guys who are (a) downvoting my original comment and (b) telling me to "educate myself" because I deigned to point out that a sample size of 31 people is low are just being petty.
They sampled 31 people. Would you accept a 31 person study over a two week period for a new pill in which "most" of the people given the pill (16 out of the 31) were now immune to any form of cancer? Or would you require more people in the study and/or longer, more rigorous testing before you took the pill yourself or gave the pill to your family?
Pointing out that there are but 31 people in the study could be considered helpful, but a basic familiarity with qualitative studies of anything that isn't a drug informs us that small sample sizes are pretty standard.
Complaining about a small sample size screams of ignorance; there's nothing wrong with shameless extroversion, thespianism, or small sample sizes. If you want to point it out and be meaningful, comment on the sample size in relation to the effect size in a way that includes statistical power. Say something. Don't just echo things you've heard other people say.
At the very least, I would require replication of the study by an independent researcher before I believed the study result as a settled conclusion about the reality of human behavior.
I know that when we run A/B tests on our website, and calculate confidence intervals, we often need thousands of conversions before we see results that we are convinced are statistically significant.