Spoilering since this is an executable. You’ll just need to trust that it’s virus-free, I guess, but it was developed by me and has been modified by no other. Since it’s a Java jar, you might have to look on the web to see how to run it, though. It might not run straight off the bat, depending on how your system is set up.
For the gender ratio, I subtracted 1 from the value since it doesn’t add anything.
Some things to note:
The formula will pick a best-fit curve regardless of whether there was actually any real correlation.
Everything is a curve, since it’s based on polynomials. This performed better (i.e., the average difference between prediction and reality is closer) than linear regression, but that doesn’t mean that everything curves all nice like it does. A flat, horizontal correlation will end up as a bump or a dip, no matter what. (See point 1).
I remember noticing, back when I developed this, that Colombia was predicted to have a far lower homicide rate than given in the original data. However, I then later saw in a different set of data - for a different year - that the homicide rate in Colombia was basically what had been predicted. I’m not sure if the year that is listed on the Wikipedia was a bad year there (gang killings?), but it does seem plausible that countries on the edge of a metric are more likely to fluctuate wildly, so you might take that into account if you try dialing in different countries.
So overall, this is possibly indicative of general ideas, probably not very good for real predictions, but also not horrible. It’s cold, reliable, math, but it’s not magic.