When the investigator is obtaining information in an undisguised manner directly from persons, the before and after control group design is inadequate. Both the experimental and control groups are apt to be influenced and in different ways, by the before measurement. To overcome these difficulties a four group – six study design is established as the ideal where there is interaction between the respondent and the questioning process. This design helps the researches measure the size of the interaction. The design can be depicted as follows:
Group 1 Group 2 Group 1 Group 2
Before measurement Yes (x1) No Yes (y1) No
Experimental variable Yes Yes No No
After measurement Yes (x2) Yes (x3) Yes (y2) Yes (y3)
Experimental group 1 and control group 1 form before and after with control group design. An additional two groups, one experimental and one control are added. Neither is measured before the experimental variable is introduced. The variable is introduced into the two experimental groups only, and all four groups are measured after. All four groups are pre-selected in such a way that they are equivalent (i.e. subjects are assigned to the four groups on a random basis). This means that the before measurements should be the same in all four groups except for random variations. It is presumed, therefore that the two before measurements will be approximately equal (x1=y1) It is then inferred that the other two groups would have shown similar measurements if they had been measured. The average of x1 and y1 is therefore taken to be the before measure of experimental group 2 and control group 2. If the before measurements had no effect on the variable being studied (e.g. purchase of dietetic colas) the two experimental groups should give the same after measurement. If the experimental variable had any influence, the results in the two experimental groups will differ significantly from the results of the two control groups.
If the before measurement does influence the test subjects directly and also interacts with the experimental variable, as is probable each of the four groups will give a different after measurements and the differences between the before and after measurements in the four cases will be the result of various factors as shown in the table on the next page. These results can be presented, as shown below, as four simultaneous equations with four unknowns. They can then be solved to obtain a value for each of the four unknown factors affecting the difference between the before and after measurements.
This four six study design may be taken as a model for marketing experiments in which data are collected from individuals in such a way that they realize it is being done. The design, however, has little practical value. The expenses on selecting four groups in such a way as to insure they are groups make this design impractical for most marketing studies. The use of inferred before measurements also creates statistical difficulties in testing the significance of results.
Despite these practical and theoretical weaknesses, the four group six study design is a useful ideal against which to compare proposed designs. Such comparisons emphasize the assumptions on which the more simple designs are based and, thus, emphasize the limitations of the data collected in the more simple designs. Fortunately, marketing experiments often permit the collection of data without the knowledge of the consumer involved (e.g. sales data) and hence the more simple design before with control group design can be used. Even more fortunately, many marketing experiments lend themselves to the more simple design after only with control group design.