QUASI-EXPERIMENTAL DESIGNS
Some people feel that if rigorous adherence to an experimental design cannot be achieved, then evaluation is not worth doing. This is shortsighted. Quasi-experimental designs can be a viable way to conduct evaluations.18 In these designs, the researcher attempts to control as many threats to validity as possible while matching evaluation research concerns to organizational constraints. Two quasi-experimental designs include the nonequivalent control group design and the time series design.
Nonequivalent Control Group Design. In a nonequivalent control group design, one cannot assume that the two groups—the control group and the group about to receive training—are equivalent. For example, if employees in one location are the trainees and employees at another location make up the control group, any number of factors may lead to nonequivalence, such as years of experience, economic conditions, and equipment.
The threats to validity most likely to affect conclusions from a nonequivalent control group design are selection-maturation interaction, testing-training inter- action, and regression effects. The burden for an evaluation researcher is to attempt to discover the factors on which the groups differ and then attempt to control for them, either statistically or in the way the experiment is conducted. A study of massed versus spaced sales training in Germany made effective use of this design to show the effectiveness of spaced practice in a banking setting.
