Double Pretest design

Explanations > Social Research > Design > Double Pretest design

Description | Example | Discussion | See also

Description

This design can be used when random selection is not possible, yet the groups selected are large enough to allow for random assignment from within them.

Start with non-random selection of two groups, one for treatment and the other as a control, then randomly sub-divide each group into two sub-groups. One sub-group from each of the treatment and control groups is given a pre-test, then the other sub-group continues with the experiment, with the treatment group being given the treatment and the post-test and the control group just being given the post-test.

In design notation, this might be written as:

N	R₁	O
N	R₁		X	O
N	R₂	O
N	R₂			O

Example

Two school classes are taking part in an educational experiment, one as the control and one as the treatment group. The classes are randomly split in half and the scores from tests are shown below.

Class	Sub-group	Pre-test	Post-test
A	A₁	12
A	A₁		20
B	B₂	11
B	B₂		13

The result shows a significant improvement in class A, and that the 'increase' of score between the two B subgroups cannot be due to priming or learning.

Discussion

A normal non-equivalent design (with assigned, rather than random, groups) may suffer from priming or learning, where the post-test result is improved or otherwise affected by having taken the pre-test.

The Double Pretest has a higher internal validity as it avoids any priming or learning effects by preventing any person taking a test twice. It assumes that with random assignment within a group, any variation in score between the pre-test and post-test can only be due to random variation. This is not quite true if the pre-test and post-test is separated by a long period in which other learning factors may occur.