- All lines are curves so stop drawing straight lines all the time. Basically, anytime you are using a linear graph, with a straight line, you can be very confident that your chart is wrong.
- Be careful with regression calculations. They are easy to do but often make up a story that is far different than reality.
- Beware of the Gamblers fallacy. Each instance has an equal chance of occurrence regardless of how many instances there have been before.
- Don't talk about percentages of numbers when the numbers might be negative. For example, percent of sales is a bad practice because it deludes the story: if sales are down 50%, is everything down 50% or is a single category doing so poorly it drags everything else down? Expenses, revenue and populations, however, are rarely negative (an expense that pays you?) so they are acceptably expressed in percentages.
- Small data sets are susceptible sample volatility. Beware.
- Inference is a tricky sport that can often lead to dangerous conclusions.
- Things are not usually equal and the inequalities often matter a lot.
- Lead with data, it will show who should win.
- Instead of asking what the chances of getting into a slot are, ask what are the chances that something in the slot is wrong. The chances you got something wrong are often the scary numbers we should be concerned about.
13 July 2016
"How Not to Be Wrong: The Power of Mathematical Thinking" by Jordan Ellenberg
It was not until I was halfway through this way thick book that I realized something important: the book's title is incredibly clever. Reading the title, I assumed that in explaining how not to be wrong that it would explain how to be correct. In reality, there is a lot of ground between not being wrong and being correct. Some interesting points: