Completely Flawed.

I concur with this comment. It seems odd not to set a confidence interval and use a hypothesis test on this data.

As far as I can tell, this article is trying to do a form of hypothesis testing. Unfortunately, it’s not doing it very well. One type of hypothesis test is to see whether the correlation coefficient is significantly different from 0, which is in fact equivalent to a t-test for most data. Unfortunately, these are not data which are amenable to a regression analysis. In any case, a proper hypothesis test requires calculating a p-value, rather than dogmatically rejecting any correlation coefficient below an arbitrary threshold of .1 (WTF? Where does this even come from? A stats book from 1988? We have easy access to computers now, you know, which let us do all these analyses directly rather than relying on cheap heuristics designed to save paper and ink…)

For the data at hand, frequency counts which can be formatted into a 2×2 table (seed/no seed vs yawn/no yawn), the correct test is clearly a chi-square test of independence. Using this test (and without applying the continuity correction), seeding and yawning are clearly not dependent on each other (X^2=.105, df=1, p=.746), which is to say that we cannot say from these data that seeding has any effect on yawning.