Saturday, May 10, 2008

Statistics Are a Gas

I went to get gas this morning ($3.79! zoinks!--but then in London it works out to something like $10 a gallon, so maybe I shouldn't complain). When I got back I thought it would be interesting (there being a definition of "interesting" I'm sure you are not familiar with) to track how much I spend on gas since I bought my Camry almost a year ago. Fortunately, it's easy to get relatively consistent series since a) I always put off getting gas until the little gas pump indicator light shows, then I fill the tank, and b) I always use the same credit card and I track all my credit card purchases in Excel because I am basically a high-tech Dickens character. And unless I am traveling, I always go to the same gas station.

The chart below shows how this worked out:
Remembering one of the chapters in Daniel Huff's classic How to Lie with Statistics, I can manipulate the interpretation of these data by simply changing the scale in Microsoft Excel. So on the one hand, I can see that my spending on gas has gone wa-a-a-a-a-y up in the past 11 months:
Whoa! It's through the roof! But not to worry: with only a few clicks, I can make my gas spending far less egregious:
Ah, there, that's better. Gas has hardly gone up at all.

I could also compare this to what I spent on gas in 1980 which was essentially nothing, since I was 13 years old then.

Just remember that when people use statistics and other mathematical legerdemain (or what Dr. Joe calls "calcubation"), they can prove just about anything they want to. Fortunately, I have nothing to prove and am just being silly.

No comments: