Treatment(s)

Two-Treatment Research

Consider a research study designed to determine the effects of damage to a particular nucleus in the brain. The researcher randomly assigns 20 rats to each of two groups. The rats in the one group have an electrode placed at the location of the nucleus of interest, and then electrical current is applied to damage that nucleus. The experimental treatment here involves everything done to these rats as part of the research—how they are housed, fed, and prepared for surgery; the placement of the electrode in the brain; the electrolytic damage to the nucleus; and so on. The other group of rats receives a different treatment—sham surgery. They are treated exactly the same as the lesioned group, with the exception that the electrical current is not applied to the electrode that is placed within the brain. Sometimes such a group is called a control group. For this group, the (control) treatment is how they are housed, fed, and prepared for surgery; the placement of the electrode in the brain; and so on, but no electrolytic damage is done to the nucleus. It is important that the two groups be equated on all aspects of their treatment with the exception of the one (or more) aspect(s) of interest to the researcher. Accordingly, when the two groups of rats are compared on some outcome variable(s) of interest, any significant differences found can be attributed to the treatment variable.

For this example, the treatment variable is “type of surgery,” and it has two values, “lesion surgery” and “sham surgery.” After the animals have recovered from the surgery, they are tested on one or more outcome variables and then the groups are compared on the outcome variable(s). If the [Page 1532]outcome variable is categorical (for example, does the rat approach or flee when a strange conspecific is presented), then the data might lend itself to a contingency table analysis (typically done with chi-square), which, if statistically significant, will lead to the conclusion that the treatment variable does affect the outcome variable. If the data are continuous and normally distributed, then a t test or ANOVA would typically be employed to determine whether the treatment variable affected the outcome variable. If the outcome variable is continuous but not normally distributed, then one would typically employ one of the several available techniques that do not require normality.

...