4 ms·

This is interesting, but I wouldn't consider it a conclusive study. According to the original paper, the sample size is only 21 (12 napping pilots, 9 no-rest pi

This is interesting, but I wouldn't consider it a conclusive study. According to the original paper, the sample size is only 21 (12 napping pilots, 9 no-rest pilots).

trentmb 13y ago
Are pilots representative of the population? You could have 1000 pilots and it still wouldn't be conclusive...
- KingMob 13y ago
  True, but this isn't necessarily a showstopper. Unless you think pilots react to sleep differently or are grossly more/less sleep-deprived than everyone else, it's not implausible to assume it generalizes. Every study has flaws, but rarely are they so awful the entire result should be discarded.
pyduan 13y ago
This is a comment that often comes up when the sample size doesn't seem very big, but keep in mind that the required sample sizes to draw statistically significant conclusions can be surprisingly small especially when the effect size is large. According to Figure 14 it seems to be the case, with the difference between the two group means being large enough to be significant at p < 0.002 (according to their numbers). A good threshold is usually considered to be < 0.05, so this is a pretty strong result. (The p-value is the probability this difference is simply a statistical fluke and not actually meaningful.) Here [1] is a discussion on how to pick a good sample size for an experiment. As you can see in the table to the right, if the effect is strong a sample size of just a few dozens can be perfectly acceptable! Here the biggest threat to the validity of these results is not the sample size, but whether one can generalize the case of the pilots in-flight (a very specific task performed by a very specific type of individuals). As trentmb says it's not clear whether pilots are very representative of the general population in this situation, although it also looks like there have been multiple studies done on the subject [2] which seem to confirm the benefits of napping as well. For what it's worth it definitely corroborates by own experience as well so I'm actually quite curious to see where Napwell will go. [1] http://en.wikipedia.org/wiki/Sample_size_determination#Required_sample_sizes_for_hypothesis_tests http://en.wikipedia.org/wiki/Sample_size_determination#Requi... [2] http://en.wikipedia.org/wiki/Power_nap#Benefits http://en.wikipedia.org/wiki/Power_nap#Benefits
coldtea 13y ago
It's not a drug trial or a voting poll. You don't need that many samples when the statistical effects are large enough.

This is interesting, but I wouldn't consider it a conclusive study. According to the original paper, the sample size is only 21 (12 napping pilots, 9 no-rest pilots).

trentmb 13y ago

Are pilots representative of the population? You could have 1000 pilots and it still wouldn't be conclusive...

KingMob 13y ago
True, but this isn't necessarily a showstopper. Unless you think pilots react to sleep differently or are grossly more/less sleep-deprived than everyone else, it's not implausible to assume it generalizes. Every study has flaws, but rarely are they so awful the entire result should be discarded.

pyduan 13y ago

This is a comment that often comes up when the sample size doesn't seem very big, but keep in mind that the required sample sizes to draw statistically significant conclusions can be surprisingly small especially when the effect size is large. According to Figure 14 it seems to be the case, with the difference between the two group means being large enough to be significant at p < 0.002 (according to their numbers). A good threshold is usually considered to be < 0.05, so this is a pretty strong result. (The p-value is the probability this difference is simply a statistical fluke and not actually meaningful.) Here [1] is a discussion on how to pick a good sample size for an experiment. As you can see in the table to the right, if the effect is strong a sample size of just a few dozens can be perfectly acceptable! Here the biggest threat to the validity of these results is not the sample size, but whether one can generalize the case of the pilots in-flight (a very specific task performed by a very specific type of individuals). As trentmb says it's not clear whether pilots are very representative of the general population in this situation, although it also looks like there have been multiple studies done on the subject [2] which seem to confirm the benefits of napping as well. For what it's worth it definitely corroborates by own experience as well so I'm actually quite curious to see where Napwell will go. [1] http://en.wikipedia.org/wiki/Sample_size_determination#Required_sample_sizes_for_hypothesis_tests http://en.wikipedia.org/wiki/Sample_size_determination#Requi... [2] http://en.wikipedia.org/wiki/Power_nap#Benefits http://en.wikipedia.org/wiki/Power_nap#Benefits

coldtea 13y ago

It's not a drug trial or a voting poll. You don't need that many samples when the statistical effects are large enough.