It's because the author is bad at interpreting statistics and doesn't recognize that a 4% difference when the average rate is closer to 18% is a way bigger difference than a 5 or even 6% difference when the average rate is closer to 40%.
The manuscript highlights the differences among the male populations are significant while those among the female populations are within the estimated confidence intervals.
Yes, but the correct interpretation of that is not that the variance is too small, but that the population is too small.
The correct conclusion to reach here (and appears to match what the study says) is that there is a huge gender gap, amongst males there is a statistically significant racial gap, and amongst females the sample size was too small to detect if a similar gap exists.
The female racial gap may not be statistically significant, so you can't say for sure if it is there, but it is possible that it is actually much more significant than the male racial gap. The article says there is "little race variation in arrest rates among females", which is not true. There is more variation than amongst the male population --it just might not be statistically significant.
At best you can say, "despite even bigger variance between races in the female arrest data, the sample of female arrests was too small for it to be conclusive".
Correct. To see the original, you can check the journal itself [0] (check for it on google scholar if this link doesn't work) or look for it at your nearest academic library, if they have it. It is published in the April issue. The conclusion was that the difference was not significant for the female population. Keep in mind that the study was based on a limited sample so care must be taken when generalizing conclusions from the starting data to the population as a whole.
This is a curious approach to the study. You'd think you'd at least consider using census data paired with arrest data. Sure there are statistical biases there, but they'd probably be smaller than the sampling error.
