A simple analysis
If there are g treatments (or repeated measures) in the data, different pairs of treatments from the g can be analysed as paired data. For example, we can calculate the differences between the two selected treatments and find a 95% confidence interval for the difference between the treatment means, as described in the previous section.
There are g(g - 1)/2 possible pairs of treatments that can be compared and this quickly becomes unmanageable if the number of treatments, g, is large. However in some data sets, we can restrict attention to a smaller number of comparisons:
If one of the g treatments is a standard or baseline treatment, we can compare the other g - 1 treatments to this baseline.
Later in this section, we will explain how to simultaneously compare all treatments, but we first show two examples in which paired confidence intervals are used.
Codeine and acupuncture for dental pain relief
In a randomised block experiment, 32 subjects were grouped into blocks of four according to an initial assessment of their tolerance to pain. Four pain relief treatments were randomly allocated to the four patients in each group and pain relief scores were recorded from each subject two hours after dental treatment.
There is a control treatment (the placebo with no codeine and inactive acupuncture points) so we will compare the three other treatments to this one.
Effect of nicotine on tics
This repeated measures experiment involved 10 patients with Tourette's syndrome, a neurological disorder that results in tics. The number of tics was recorded during an initial 30-minute period (baseline), then during 30 minutes of chewing nicotine gum (containing 2 mg nicotine), from 0 to 30 minutes after chewing, and finally from 30 to 60 minutes after chewing.
There is again a baseline measurement for each patient against which the other measurements can be compared.