Plot your data.

Writers of statistics textbooks tend to copy other textooks rather than draw on experience. This leaves a serious ap: Techniques that are highly useful in practice are not aught. With this series of columns I am trying to fill that ap. The first column [1] pointed out that doing something imperfect) is better than doing nothing. The second [2] was bout the value of transforming your data. A third neglected lesson about data analysis is plot your ata. More precisely, make all reasonable graphs of your ata. Make a histogram of every measurement (to see its istribution), plot every measurement against its date, and lot every measurement against every other measurement. his is a good way to generate ideas. To read almost any statistics textbook, even the best e.g., Box et al. [3]), you’d think science was all about esting ideas. It isn’t. Where do the tested ideas come from? dea generation, which these books ignore, is just as imortant as idea testing. One of the best ways to generate new deas worth testing, I have found, is to make many graphs of y data. It is like searching for buried treasure. New ideas worth esting are very valuable but hard to find. Only a tiny raction (1%?) of the graphs I’ve made led to new ideas but ome of those ideas had a big effect. My graphs generated new ideas in two ways. 1) Causalty. The graph suggested a cause–effect relation I hadn’t hought of. 2) Simplicity. Something turned out to be simler than expected. Here are examples.