Ravenwood - 08/09/02 10:49 PM
I had to read this article in my Popular Science magazine three times, because I couldn't believe that Popular Science would fall for such a ruse. The hypothesis is that child suicides by gun are higher in states where are large percentage of households own guns, while non gun suicides remained almost the same. This leads researchers to believe that more guns yields more suicides.
The problem is that actual numbers of the number of households that own guns is very difficult to calculate. The Harvard Injury Control Research Center who performed the study chose to use a proxy called Cook's index rather than actual statistical sampling. In other words, to support their claim, researchers extrapolated the percentage of households with guns from among other things, the number of gun suicides in the state. The circular logic is astounding. The study concludes that gun suicides are higher based on the number of households that own guns, BUT the number of households that own guns is calculated directly from the number of gun suicides.
Researchers also failed to use collectively exhaustive outcomes, by assuming up front that the number of suicides is directly proportional to the prevalence of guns. In simple terms, this means that the study categorized states into two groups: states with a low proportion of gun suicides and low prevalence of guns and states with a high proportion of suicides and high prevalence of guns. In fact there are two other possibilities that cannot be ignored, states with a low proportion of gun suicides and high prevalence of guns, and states with high proportion of suicides and low prevalence of guns. Ignoring these possibilities guaranteed that the research would show a linear directly proportional relationship between suicides and gun ownership.
It is sad that Popular Science seemed to buy into it. I also notice that this is the second anti-gun article in three months, which makes me wonder if the ideological views of the editors are starting to shine through.
Click here for a greater analysis of the statistics. This is an old study, and it has been debunked by several organizations.
(c) Ravenwood and Associates, 1990 - 2014