When you look at the distributions of your variables (often through graphs of the data like the histogram in Chapter 7), you may see that some of your variables are oddly distributed. This, in turn, may need extra or different statistical analyses.

Say that you are a banking customer relationship manager who wants to analyze customer satisfaction feedback made at the time of a transaction, and that the customers are only asked one question about service with replies on a scale: “Very satisfied,” “Satisfied,” “OK with our service,” “A bit unsatisfied,” “Very unsatisfied.”

If you take an exploratory peek at the distribution of the data you may find that it has a very bimodal distribution like that in Chapter 7: people tend to answer either “Very satisfied” or “Very unsatisfied” and seldom choose an option in-between. If you see this, you should make sure the analysts doing statistics account for whether and how they need to deal with the strange distribution. You will be surprised how often you will avoid analyst error or laziness.

If you have multi-item scales, then the analysis process includes decisions about how to deal with them. The wrong decisions about multi-item scales can lead to poor final statistics. The next section discusses major statistics and statistical processes surrounding multi-item scales.