My last post Mass Shootings Unleash the Emotive, Innumerate Hounds addressed some concerning statements made in The New York Times article, What Explains US Mass Shootings? International Comparisons Suggest an Answer. But that article is full of more anti-pattern gems showing exactly how one should not make international comparisons.

The Mental Health Decoy Defence

If mental health made the difference, then data would show that Americans have more mental health problems than do people in other countries with fewer mass shootings.

Not sure how one gets there. All you need to do is to show that people who commit mass shootings tend to have mental health issues. How many of the mass shooters had mental health problems so far?

If you have data showing how Americans have more mental health problems than other people, you just have a prior probability. You can't compare this to people in other countries with fewer mass shootings, you can only compare this to the mental health data in other countries, provided that you're using an appropriate metric of incidence rates.

But you can compare mass shooting incidence rates of different countries directly, and see how many of the different perpetrators had mental health issues. I would venture a rough guess that most of those incidents are split between perpetrators with mental health issues, people with some very strongly-held religious convictions, and people with strongly held political convictions. No comment regarding correlation there.

A 2015 study estimated that only 4 percent of American gun deaths could be attributed to mental health issues.

A gun death is not necessarily a mass shooting. In fact, 4% of overall gun deaths may very well be more than all the gun deaths due to mass shootings, given that mass shootings are relatively rare, their total gun deaths is comparatively small and a considerable amount of gun deaths is due to suicide.

So what does this 2015 study, Mental illness and reduction of gun violence and suicide: bringing epidemiologic research to policy, claim? Does it suggest that mental health isn't really related to gun deaths?

Media accounts of mass shootings by disturbed individuals galvanize public attention and reinforce popular belief that mental illness often results in violence. Epidemiologic studies show that the large majority of people with serious mental illnesses are never violent. However, mental illness is strongly associated with increased risk of suicide, which accounts for over half of US firearms-related fatalities.

The referenced study doesn't appear to make any claims about 4 per cent of gun deaths. In fact, it shows that over half of American gun deaths are attributable to suicide, which means over half of US gun deaths are directly related to mental health issues. It claims exactly the opposite of what The New York Times's claims, but it does caution against the notion that mentally ill people are necessarily violent.

You Should Be Comparing Incidence Rates, Not Raw Numbers

Sceptics of gun control sometimes point to a 2016 study. From 2000 and 2014, it found, the United States death rate by mass shooting was 1.5 per one million people. The rate was 1.7 in Switzerland and 3.4 in Finland, suggesting American mass shootings were not actually so common.
....
But the same study found that the United States had 133 mass shootings. Finland had only two, which killed 18 people, and Switzerland had one, which killed 14. In short, isolated incidents.

So the author takes an incidence rate and converts it back to raw numbers, thereby making the raw numbers incomparable in any meaningful way? Then claims this trumps incidence rates, which can in fact be compared?

I quite like the 2015 study referenced in The New York Times. It concludes with:

Policymaking at the interface of gun violence prevention and mental illness should be based on epidemiologic data concerning risk to improve the effectiveness, feasibility, and fairness of policy initiatives.

Excellent. So let's use epidemiological data instead of massaging epidemiological data back into useless raw numbers? Otherwise, just pick the best memes and let them decide the issue?