All these epidemiologically based studies (like Kellerman’s) suffer from the same basic defect: they not only conclude that correlation equals causation, but they assume a direction for the causation that has no real support. In fact, causation in the other direction is much more likely.
Specifically, they assume that possessing a gun somehow makes one more likely to get shot. It is much more logical to think that the fear of getting shot makes one more likely to possess a gun. It should therefore not be surprising that there exists the kind of correlation that Kellerman reports.
Using Kellerman’s logic, one could say that since many people die in hospitals, sick people should avoid hospitals at all costs.
Also, if I remember correctly, in Kellerman’s original study a large proportion of the subjects were either drug users or drug dealers. Participants in the illegal drug trade would likely have a greater-than-normal fear of being shot.
Zeko
The infamous "43 to 1" statistic which was spawned by this 'study' postulates that a gun in the home is 43 times more likely to kill the owner or a family member than to kill an intruder. 37 of those 43 instances were suicides, so the people who died under those circumstances chose to kill themselves and would have done so regardless of whether a gun was available.
That still leaves 6 to 1 where someone in the home is supposedly more likely to be killed than in intruder. The reason I underlined the word killed is because any instances where brandishing the weapon was enough to scare off an intruder, or firing the weapon resulted in either no injury or a non-lethal injury to the criminal was not counted as a defensive use of the firearm.
Also; in domestic disputes where a woman defends herself from an abusive spouse or partner, a self-defense death was counted as a 'family member being killed' for the purposes of the study.
Perfect example of a study being manipulated to support a predetermined conclusion.