Data, Variables, and the Statistician's Toolkit
The toolkit of the program. What data are and the kinds of variables, population and sample, parameter and statistic, the summation sign and the mean, proportions, percentages, and rates, relative against absolute change, powers of ten and the logarithmic axis, the rules of an honest graph, and how data are collected, with a short history of counting people from the Bills of Mortality to Hollerith's machine.
1 Introduction
Nothing in the world comes in one size. Two patients who take the same tablet sleep for different lengths of time, two fields sown with the same seed give different harvests, and two coins tossed by the same hand land differently. Whoever wants to say something true about patients, fields, or coins has to say it about many of them at once, and nobody can look at all of them. The farmer sees her own fields, the doctor sees his own patients, and the pollster who wants to know how a country will vote can ask a thousand people out of fifty million. Every claim about the world rests on a part of the world, and the part is never the whole.
Statistics is the discipline of reasoning carefully from the part to the whole. It began with people who counted. In 1662 a London shopkeeper named John Graunt (1620-1674) collected the weekly lists of burials that the parish clerks of the city had printed for decades, added them up over sixty years, and noticed things nobody had seen before, such as the fact that slightly more boys than girls are born every single year, or that the plague returned in waves. Two centuries later Florence Nightingale (1820-1910) came home from the Crimean War with the knowledge that far more of her soldiers had died of disease in the hospitals than of wounds on the battlefield. She turned the counts into a diagram, the one on the title page of this unit, and sent it to the Queen, because she had understood that a picture of the numbers could change an army where a table could not. In 1908 a chemist at the Guinness brewery in Dublin, William Sealy Gosset (1876-1937), who was allowed to publish only under the pen name "Student", worked out how much a sample of four or five measurements can tell about the whole batch of barley it came from, and in doing so he gave medicine and agriculture the method by which small experiments are judged to this day.
The same reasoning now decides whether a new vaccine works, because the only way to find out is to give it to thousands of people and a harmless substitute to thousands of others, and then to compare how many in each group fall ill. It decides, on election night, whether a candidate can be declared the winner from a fraction of the count. It decides which of ten thousand genes is connected with a disease and whether the warming of the planet is a trend or a fluctuation. In each case someone has a part and wants to speak about the whole, and in each case the honest answer has two halves: the best estimate, and a statement of how far it could be wrong.
That second half is what makes a statistician different from a calculator. A calculator divides 7.5 by 10 and reports 0.75. A statistician asks where the ten numbers came from, whether ten more would have given 0.75 again, and how much the answer deserves to be trusted. Learning to ask these questions, and to answer them with numbers, is what this masterclass is about. This first unit assembles the tools: what data are, how they are summarized with sums and proportions, how percentages and rates are read and misread, why so many graphs in this subject have a logarithmic axis, and how a graph is drawn so that it tells the truth. Everything in it will be used in every later unit.