Today is World Statistics Day, celebrated for the first time on 20 October 2010 worldwide in accordance with a declaration by the United Nations Statistical Commission. The Royal Statistical Society in the UK launched its getstats “statistical literacy” campaign to open the celebrations at 20:10 on 20.10.2010. I don’t want to get too technical here; I have plenty of experience watching people’s eyes glaze over when I start talking mathematically. The fact is that statistical analysis can be extremely complex, but the foundational ideas are really easy to grasp. For a set of funny and informative videos I suggest you go here https://worldstatisticsday.org/ Here I’d like to do a couple of things, namely talk about the handling and presentation of statistics, and have a little fun.
Like it or not, statistics rule a big chunk of our lives. I’m a social scientist so statistics are a big part of my professional life, even though a lot of my writing is math-free. “Proof” of assertions concerning social life hinges on good statistical data. You may think something about social life is obviously true, but you need statistical data at your back. Three things are important – (1) data do not speak for themselves, (2) proper presentation of data is vital and (3) data are only as good as their method of collection.
The first point ought to be self evident, but often is not. You cannot show me some data and assume I will see in them what you see. Suppose you show me a graph of rising fuel costs over the past decade. What should I do with it? Does it matter to me? If it does matter to me, how does it matter? Is it a good thing or a bad thing, for me, or in general? It does not speak for itself. Maybe I own a factory and rising fuel costs are eating into my profits. Maybe I am a worker whose salary increases have not kept pace with inflation, so I am having to cut back on non-essentials. Maybe I am a hermit living in a remote cave with no need to buy fuel. Context matters in interpreting statistics.
The second point can also be overlooked. In the 19th century Florence Nightingale discovered that in military hospitals in the Crimea and elsewhere, a great many more soldiers died from preventable diseases than from war wounds. She believed that better sanitation in the hospitals was the answer but she needed to convince bone headed politicians to vote for increased funding. To do so she felt that if the data were graphically presented they would be more understandable than tables and spreadsheets. So she created a type of pie chart now sometimes called the Nightingale Rose – shown here (click to enlarge):
It was effective, although it’s debatable whether this chart was more effective than a simple bar graph as shown here (click to enlarge):
You decide. At very least you understand the importance of method of presentation.
The third point can also be overlooked very easily. Probably everyone understands that when you are conducting a survey, the size of your sample and the nature of people in the sample are critical issues. You can’t get a meaningful picture of racism in the U.S. by polling a small group of white people all living in one state. You have to have a large, widely distributed sample of people from all walks of life and all ethnicities. But the quality of the data also depends on the questions asked and the responses allowed. Obviously you can’t just bluntly ask, “are you a racist?” You have to decide what questions will get at the heart of the matter, and that is far from easy. You also have to contend with the fact that many people who answer surveys answer according to their ideal self image, and not necessarily according to the truth.
Here now is a little gallery of amusing statistical charts:
As long-time readers know, I like to cook by the seat of my pants most of the time, and it’s something of a strain to come up with precise measurements and instructions. So here is my heuristic/statistical recipe for a pear and passionfruit crumble I made yesterday using percentages. It’s pretty close to how I actually think when I cook.
© Pear and Passionfruit Crumble
With a fruit crumble the correct ratio of fruit to crumble topping is very important. By eye I would say my crumble is 35% topping and 65% filling. Some people may like to have more fruit. The 35% topping is divided thus: 10% rolled oats, 10% plain flour, 10% granulated sugar, and 5% butter, or a ratio of 2:2:2:1. Put the oats, flour, and sugar in a mixing bowl and stir a little until they are mixed. Make sure the butter is very cold and cut it into the smallest pieces you can. Rub the butter into the dry ingredients with your fingers so that the mixture is reasonably homogenous. Set aside.
For the 65% fruit mix Use about 60% pears and 5% passionfruit. These days when I make crumbles I don’t peel the fruit. They have an earthier taste unpeeled. Cut the tops and tails off the pears, then slice downwards to separate the meat from the core. Discard the core and slice the meat thickly. Put the pear slices into a baking dish and sprinkle with sugar. Cut the passionfruit in half and scrape the inside pulp on to the pears. Toss with a wooden spoon.
Pour the crumble topping over the fruit and spread it evenly. Tamp down the top a little to compress the crumble a little but not too firmly. Bake in a 400°F oven for about 45 minutes, or until the top is mottled golden-brown. Serve hot or cold, plain or with custard, whipped cream, or ice cream.