Experimental Design
Experimental design requires a defined control group. For example, the investigator recruits 200 participants for her study. One hundred participants were randomly assigned to the intervention group. The remaining 100 participants will be assigned to a control group, therefore not receiving the intervention.
All experimental designs include the following properties:
Manipulation Something is “done to” the subjects, such as prescribed a new medication or a different method of teaching.
Control There is a group that is untouched and is used for comparison purposes. The control group would not receive the new medication or would be taught according to usual practice.
Randomization Subjects are assigned to a group using chance methods. Chance methods include coin toss and the use of a random number generator where subjects with an even number are assigned to one group and subjects with an odd number are assigned to the other group.
Examples of experimental design
Posttest-only: Data are only collected after the intervention. Comparison occurs between the manipulation and control group. Example: Average Weight loss between a group of subjects on a low carbohydrate vegan diet versus a group of subjects who eat an unrestricted vegan diet.
Pretest-Posttest: Compare depression scores at the start of the study for both groups. The intervention group will be required to participate in Cross Fit classes three days a week. Measure depression scores between the groups after one month.
Crossover: Both groups receive the intervention but there is a washout period between interventions. This design is a within group comparison.
Statistical analysis
Any statistical tests that compare groups, such as independent T tests, chi square tests for categorical variables (examples gender, race, age group, geography). In the case of a crossover design, the statistical tests will also need to measure within group differences; ANOVA (analysis of variance) is one such test.
Strengths and limitations
A randomized control trial is considered the “gold standard” in research. The conditions are tightly controlled so it is easy to isolate the effects of the manipulation, however, that is also the problem. The research occurs in isolation from the real world. It is also difficult to recruit a sufficient number of subjects to have an adequate sample size.