Thanks for a great article! We are constantly barraged with "statisitcs" and "conclusions" in the media that are meaningless. An example here in Wisconsin: The governor recently signed into law a 55 mph speed limit at night for snowmobile trails to reduce snowmobile fatalities. The state has been running a trial of the speed limit for three years. The governor signed the law "because it worked." But did it? I collected the fatality data back to 1992 and performed an analysis similar to the one in your article. Guess what??? There is NO statistical evidence that the 55 mph speed limit has reduced snowmobile fatalities!!! While collecting the historical data, I found articles over the years that tried to explain any "down-tick" in the data from year to year. Two "down-ticks" in a row make for a "trend" and celebrations of what the state is doing to reduce snowmobile fatalities. They never seem to explain the increase in fatalities following the two down-ticks. One of the most interesting comments: "After a year of high fatalities, the following year the number goes down because snowmobilers are more aware and become more careful."
A Process Behavior Chart of the data (1992 - present) shows the number of snowmobile fatalities to be stable. However, there are more questions to be asked. What about the technology changes in snowmobiles and safety devices? What about changing law enforcement policies? What about the number of snowmobilers on the trails? What about the "causes" of each fatality (speed, alcohol, darkness, etc.) There are many other questions to be asked before conclusions can be drawn. Torture the data untils it gives you the answers you want. It remindes me of the old saying "The figures don't lie, but liars figure."