Saturday, July 2, 2016

On Data and Science


The irrepressible Justicar recently exposed a fraudulent study published in the Journal of American Medical Associates linking Australia’s 1996 gun-restriction legislation to decreases in mass shootings, in which “mass shootings” are bizarrely and uniquely defined as having 5+ fatalities, rather than the usual 4+ used by the FBI and virtually every other serious organization studying the issue. This allows the study authors to tap in a triumphal “0” in the 1997-2013 mass shootings list, and lends credibility to the juxtaposed 1979-1996 firearm homicide mean (0.56/100,000) and 1996-2013 firearm homicide mean (0.2/100,000).
Off the bat, the minute scale of the increased safety achieved should be noted, even if the data istaken at face value. This is like banning vending machines in order to save the 13 people killed per year in the United States, on average. At some point, even convenience is worth a few lives on scale. How much more important than vending machines are firearms? They are not merely a hobby, a source of personal protection, and provider of food, but are enshrined in our Constitution because they are a doomsday provision against a tyrannical government. (No, superior military technology does not make defense against the government moot, or else we would not still be fighting illiterate shepherds with Kalashnikovs in the hills of Afghanistan).
But the data shouldn’t be taken at face value. Let’s break the firearm homicide rate down slightly more honestly:
1979-1981 – 0.65/100,000
1982-1984 – 0.68/100,000
1985-1987 – 0.68/100,000
1988-1990 – 0.46/100,000
1991-1993  – 0.47/100,000
1994-1996 – 0.44/100,000
1997-1999 – 0.32/100,000
2000-2002 – 0.27/100,000
2003-2005 – 0.14/100,000
2006-2008 – 0.16/100,000
2009-2011 – 0.16/100,000
2012-20013 – 0.17/100,000
Wouldn’t you know it, there was a downward trend prior to the enacted legislation.
It is tempting to see the 0.0001% increase in the rate of the preexisting decline, which does appear to be possibly attributable to firearm restrictions. This would, of course, involve trusting the data itself, provided by these dishonest scholars. I’m using it above for convenience and demonstration but otherwise wouldn’t bet my life on its veracity. But this itself excludes the very important issue of non-firearm related homicides. If you’re interested, feel free to study the data for yourself. Suffice to say, firearm restrictions did not meaningfully reduce homicide generally, certainly not beneath the downward trend it was already on. Those too eager to jump from data correlation to causation with the immediate decline might also find themselves in the awkward position of trying to explain a suicide spike in 1997-1998, immediately after the 1996 gun bill. I’m not saying this is in any way related to the gun restriction bill, of course (unless you believe that the gun bill reduced rates of violence, in which case, stop hating depressed people, you murderous psycho). I am simply saying that when culture, pathology, violence, politics, income, and happiness are influenced by more factors than we can possibly account for, and when large, preexisting historical trends are already at work, it is very easy to manipulate statistics and “science” to fit one’s own position.
This is not an argument against science or statistics, for the record. On the contrary, I am in fact making an argument for science, and, a bit more begrudgingly, statistics. What I am arguing against, however, is the tendency for people to allude to other people’s conclusions allegedly based upon science or “hard data” instead of actually making an argument (these people are, almost without exception, never scientists or statisticians themselves). The argument is the essence of science. The reason that scientists follow the “scientific method” is not because God came down from the mountain and told Moses “thou shalt divide thine research subjects into two categories, and thine shalt name the first ‘control’…” The scientific method has evolved into its current form because the results are (ideally) very high quality clay with which a scientist can form a robust argument. The possession of the clay, however, does not in any way relieve the scientist or the ideological champion from the responsibility of actually making the argument. And once the argument is made, it is always open to criticism and rebuttal. The strength of an argument is its ability to withstand this inevitable and never ending scrutiny, and the moment a conclusion is held to be above challenge, it is to that degree not a scientific conclusion any more, but an ideological one.
This means that appeals to science and data are, ironically, unscientific. When an actual scientist is asked a question challenging his own beliefs within his field, what you will almost always see is an argument. He will offer an explanation that utilizes the data and research, of course, but he doesn’t just say “here’s the data” or “I can read a graph.” Those who appeal to science without bothering to make the argument don’t understand the nature of science, let alone the science to which they are referring.

No comments:

Post a Comment