3 ms·
75 subjects divided into 4 groups, so about 19 people per group. Suppose the study was studying the effect of coin tosses per group. The standard deviation is
by jpmattia 11y ago
75 subjects divided into 4 groups, so about 19 people per group.
Suppose the study was studying the effect of coin tosses per group. The standard deviation is sqrt(19)/19 or about 23%. And that is only one sigma.
So we might also conclude that those who flip heads are <insert effect here> than those who flip tails. Is anyone really surprised that the reproducibility study [1] found reproducibility comparable that of a coin toss (to about one sigma)?
[1] http://www.nature.com/news/over-half-of-psychology-studies-fail-reproducibility-test-1.18248 http://www.nature.com/news/over-half-of-psychology-studies-f...
- epistasis 11y agoThe std. dev. of a binomial is sqrt(n p (1-p)). So for 19 coin flips that's ~2.2, or 11%, not 23%. Plus your setup is not relevant to the setup that they performed (check figure 3). I am not so surprised that many studies are not replicated, but I also am not surprised when critical internet comments are weaker than the analysis they criticize.
- jpmattia 11y ago> The std. dev. of a binomial is sqrt(n p (1-p)). So for 19 coin flips that's ~2.2, or 11%, not 23%. You are correct. The usual way in which std dev is applied is ± sigma accounts for about 68% of the trials, which is 2 std deviations, and is what I had in mind. So for 19 people flipping a coin, you'll have a range of about 23% variation. That factor of 2 does not change the message.