True Experimental Design

Random Assignment and Causality

For randomized experiments, the use of random assignment is important because it facilitates causal inference. For example, randomization ensures that the experimental units’ treatment condition is not confounded by an alternative cause or systematically introduced difference between conditions. Randomization, therefore, reduces potential threats to experimental validity by dispersing these threats randomly across experimental groups to minimize group differences before treatment begins. Randomization allows the researcher to know how groups were assigned and has the advantage that the selection process can be modeled accurately. Random assignment also allows for the computation of valid estimates of error variance that are orthogonal to treatment conditions. Each advantage captures different aspects of the benefits of randomization. Random assignment is the only design feature that accomplishes all the above in the same experiment, and it does so reliably while minimizing threats to internal validity.

Designs Using Random Assignment

This section presents various randomized experimental designs. The following designs are among the most commonly used in field and treatment research and are basic designs from which more complex designs can be built. With each design, key features include control and manipulation over the timing, intensity, and duration of experimental variables. The basic design of a randomized experiment requires a minimum of two conditions, with random assignment of experimental units to treatment conditions, followed by posttest assessment of units. Herein, the letter “R” indicates that the group on a given line is formed by random assignment. Although random assignment R is placed at the beginning of each line, in practice, [Page 1544]randomization often occurs before or after pretests. “X” is the treatment and “O” is the observation. The basic design can be depicted

...