So my last analysis on this garnered a fair amount of positive attention, so I decided to do a second analysis. And probably continual analyses on this topic.
The background here is a Kansas statistician is currently claiming that there is statistical evidence for pro-Republican election fraud in the Kansas 2014 election cycle. I'm quite a bit more cautious about what I say than this statistician, but nonetheless I think it's an interesting concept, and one worth looking into from a statistical perspective.
For my second analysis I took into account other factors that are specific to individual counties to determine if the pattern would still hold up. I was actually expecting the potential-fraud pattern to disintegrate under this second layer of scrutiny, but the relationship actually became stronger.
After a career where I do quite a bit of fraud detection, I'm fairly convinced that we'll find a logical, demographic explanation for this pattern, but will continue analyzing until that path is exhausted.
Here's a link to my new analysis.
And here's a link to my original analysis.