Received a response from the reporter at KETV:
I have not replied to this yet, but I am not really happy with the reply nor the original quote, especially if this is the same study you guys mentioned earlier with a whopping 16K participants.
It is this poorly-done study:
https://www.dropbox.com/s/7vrou1xas4tiyb9/Firearms%20and%20Alcohol.pdf?dl=0The reason the data is only from those two years (1996 and 1997) and only from those few states, and only those few people---is that those are pretty much the only years in which firearms data AND alcohol data were taken from a larger enough group for Wintemute to badly generalize results. After 1997, firearms data was not taken, if I recall correctly (due to a number of issues with the accuracy of the data), and many states did NOT take that data set anyway.
I note that as normal, Wintemute's results bearing only passing resemblance to conclusions actually supported by the data, and as ALSO is normal, he still can't do math.
His Table 1 N percentages are only correct because his first column % are based on the total number of participants, but the REST of his numbers are based only on the number of people who had guns in the first place, and his odds ratios are completely unnecessary as this data is merely the raw numbers compared to a semi-randomly chosen referent that has nothing to do with anything. It also attempts to make conclusions based on numbers as low as single digits, yet forms "odds ratios" for risk assessment even though at least once the referent has a data value of zero.
Um, it doesn't work that way.
And that's just the demographic and raw data section.
Note: In there, while his first column of data is based on a sample size of 15-16K, the REST of his calculations are based on a sample size one third of that.
Looking at Table 2, it gets worse---the sample he is basing his numbers on 1) varies in size by demographic of choice, and 2) includes extremely TINY sample populations, from which conclusions should not be reached.
I'm not sure what statistical software package he used to calculate his odds ratios in tables 2 and 3 (or in 1, really, but it is less important there because it is completely meaningless in table 1 compared to the others) but it certainly doesn't match any I've ever seen. Managing to get an odds ratio of 1.66 (1.54 to 1.80) when comparing a prevalence of 46.8% to 59.4% requires New Math. The rest of his odds ratios are similar.
In the multivariate analysis (which I assume he calculated no better than anything prior to it) he even says: "Firearms owners who engaged in firearm-related risk behaviours were again generally more likely than others to report alcohol-related risk behaviours, and were in all cases more likely to do so that were persons with no firearms at home."
...and yet, this is blantantly untrue. He considers carrying a firearm for protection and confronting someone else with a gun as firearm risk behaviors. And yet, in every single one of those cases, firearms owners confidence intervals showed NO statistical significance. Every single CI contained 1.0 as a possible value--for drinking ANY alcohol, >5 drinks, drinking and driving, and >60 drinks a month.
Meaning: NO statistical significance to ANY of that.
In other words, in terms of the specific things that we are talking about here (CCW folks)---even THIS incredibly biased study can't show any particular increase in alcoholic behavior among people who carry firearms for self-protection, or have had to confront someone else with a firearm.
And that completely leaves out the fact that this study does not even attempt to study the difference between legal gun ownership, and illegal gun ownership. Plus, "respondents" could be of ages 18 on up, so many of these aren't even in the right age group to drink legally OR own handguns legally. (You'll note he put 18-24 as a demographic, instead of 18-20 and 21-24.) And of course this study is about
any firearm, so an 18-year-old with a hunting shotgun is treated the same as a 30-year-old with a CCW.
His study is full of complete fail, and STILL can't actually support a contention that CCW holders are more prone to alcohol abuse.
Sorry about any typos.
