Statistics for Business: Decision Making and Analysis, 3rd Edition

17.7 Paired Comparisons

Comparison arises in many situations, and not all of these involve comparing the means in two independent samples. Data sometimes provide two measurements for each subject: paired data. For inference with paired data, we confine attention to confidence intervals.

Paired comparisons are useful because they isolate the effects of a treatment. One of the best ways to compare strategies is to put them in a head-to-head contest. For example, rather than have one group of people rate the taste of Pepsi and a different group rate the taste of Coke, we obtain a better idea of preferences by having the same people taste and rate both colas. This head-to-head approach is called a paired comparison of the treatments. It is said to ...