Robert Seeberger wrote:
...
> Evaluating the "43 times" fallacy
...a study by Arthur Kellermann and Donald Reay published in the
> June 12, 1986 issue of New England Journal of Medicine (v. 314, n. 24, p.
> 1557-60) which concluded that a firearm in the home is "43 times more
> likely" to be used to kill a member of the household than to kill a criminal
> intruder.
Most of the criticisms are valid, but there are a couple of
flaws. (I've snipped all but the flaws.)
...
> How many successful self-defense events do not result in death of the
> criminal? An analysis by Gary Kleck and Marc Gertz (Journal of Criminal Law
> and Criminology, v. 86 n.1 [Fall 1995]) of successful defensive uses of
> firearms against criminal attack concluded that the criminal is killed in
> only one case in approximately every one thousand attacks.
But this isn't fair either, since the intent of the criminal is
unknown. The factor of 1000 is used as if all of these were prevented
homicides. A large fraction were probably "prevented burglaries",
which should not be counted as high as human life. (Possessing a gun
would have to foil MANY burglaries for that to be worth a sizable
risk of killing a family member!)
...
> "Reverse causation" is a significant factor that does not lend itself to
> quantitative evaluation, although it surely accounts for a substantial
> number of additional homicides in the home. A person, such as a drug dealer,
> who is in fear for his life, will be more likely to have a firearm in his
> home than will an ordinary person. Put another way, if a person fears death
> he might arm himself and at the same time be at greater risk of being
> murdered. Thus Kellermann's correlation is strongly skewed away from normal
> defensive uses of firearms. His conclusion is thus no more valid than a
> finding that because fat people are more likely to have diet foods in their
> refrigerators we can conclude that diet foods "cause" obesity, or that
> because so many people die in hospitals we should conclude that hospitals
> "cause" premature death. Reverse causation thus further lowers the 0.006
> value, but by an unknown amount.
This is often called a "confounding variable", one factor that
increases the likelihood of both the "cause" (explanatory) and the
"effect" (response) variables in a study. They seem to be proposing
"fear of death by homicide" as a confounding variable, but it is
not stated very clearly.
One can successfully argue for some connection here.
Certainly people at high risk of being killed by homicide tend to
know this. And if one is "afraid of homicide", one is more likely
to shoot people without carefully verifying they are strangers,
leading to more accidental killings of family members.
But it doesn't seem to me to be a very strong effect, and
it could well be countered by people in an armed household knowing
enough not to do things like "climb in the window when you forget
your keys, rather than knock and wake everybody up".
---David
Where was this when I was still teaching my Statistics class?
_______________________________________________
http://www.mccmedia.com/mailman/listinfo/brin-l
