From: "Julius Chang" <[p 00302] at [psilink.com]>
Newsgroups: talk.politics.guns
Subject: Re: NEJM Study: Dumb question
Date: Sun, 24 Oct 93 21:56:51 -0500

Steve Holland <[gila 005] at [uabdpo.dpo.uab.edu]>
>Yes, a good point.  The assumption that you criticise is that the
>purchaser/owner was successful in making the defense, but it looks
>like a homicide anyway.  I think in the study such cases were thrown
>out, but I am not sure.  11 cases of justifiable homicide were excluded.

	Kellerman et al. never refer to "justifiable homicide", only
	to "excusable homicide."  No doubt that they are wrong since
	they claim that four of the 15 cases were law enforcement
	killings and some of the other 11 were in "self-defense."
	Self-defense should be a justifiable act, not an excusable one.

>I think a scenario like above would have been excluded.  
>

	I don't see this in the paper.  There were 15 cases of what
	the authors call "legally excusable homicides" (Table 1).
	On the same page, second column, second paragraph, they
	say that four victims were shot by law enforcement and the
	other 11 were "killed by another member of the household
	or a private citizen acting in self-defense."

	First, there is a legal difference between "excusable
	homicide" and justifiable homicide."  It appears that
	Kellerman et al. do not know the difference.  Unfamiliarity
	with the legal and law enforcement definitions of the
	data collection categories does not inspire confidence in
	the accuracy of their criminological data.  This goes to
	the point of the definition of an "acquaintance."  The authors
	have not proven that they know what it is and all indications
	are that they do not.

	When I further see that the authors try to support their
	findings by referencing discredited research (see page 1090,
	second column, first paragraph, references 16 and 17),
	the credibility of Kellerman et al. plummets.  It indicates
	that Kellerman and Reay have learned nothing in the intervening
	five years since their Seattle/Vancouver study.

	Second, I see no mention of excluding these 15 cases
	because they are "excusable."  Even if 11 cases were
	excluded as you claim, that suggests that the four law
	enforcement shootings WERE INCLUDED (maybe).  And those
	undoubtably are justifiable, not excusable.

	The final analysis in Table 4 is based on 316 matched
	pairs.  There were 1860 homicides during the study period.
	444 occurred in the home (focus of this research).  The
	authors excluded "24 cases for various reasons" (never
	clarified).  Down to 420 now.

	Out of the 420, matched controls (based on age, sex and
	race only) were obtained for 388.

	The final "logistic regression analysis" was conducted on
	316 of the pairs because "72 pairs with missing data on any
	of the six variables of interest were excluded from this
	analysis."  "Conditional logistic regression analysis requires
	that data on all the variables of interest be available for
	both case subjects and their matched controls."

	Thus, some or all of the excusable homicide cases may have
	been excluded, but it is never made clear.

	Based on simple percentages, 316/420*15=11 legally excusable
	cases may have been INCLUDED.  At worst, 316/420*4=3 cases
	may have been included.

	This paper is extremely fishy since a lot data is thrown out
	and only vaguely justified by the authors.  When this amount
	of data is excluded and the authors make no analysis of the
	effect of non-respondents, one wonders about the meaning of
	their 95% confidence limits (especially since they don't even
	have any clue on the statistical distribution of their raw
	data).

	If you were a referee for this paper, would you rate it
	"publish with no changes", "publish with optional suggested
	changes, "publish with mandatory changes", or "not
	worthy of publishing"?

	-Julius