Non-equivalent Dependent Variables design

Explanations > Social Research > Design > Non-equivalent Dependent Variables design

Description | Example | Discussion | See also

Description

Use this test when it is suspected that the treatment group may be affected by factors outside the treatment topic (or that unwanted 'leakage' of the treatment is occurring).

This design uses only a single group, but applies two tests to them, covering separate topics. The treatment is applied only with regard to one of these topics. Thus all people in the group take two pre-tests and two post-tests.

In design notation, this can be written as:

N	O₁	X	O₁
N	O₂		O₂

Example

A History teaching method is being tested in a school with a single class. Standardized tests are used in both History and English.

The score below shows that whilst the History score increases, the English score has not increased. It thus may be concluded that any change in English teaching or other factors has not affected the increase in the History score.

	Pre-test	Post-test
English	12	13
History	6	12

Discussion

In this design, the treatment group is effectively its own control group. The actual control, though, is the additional variable. Note that this 'control' variable (English, in the above example) should be similar enough to the treatment subject to be affected by the same sort of factors.

This design can be used in two ways. First, it can check for the possibility of changes elsewhere contaminating treatment results, for example where a change in teaching in another subject contributes to changes in results in the treatment subject.

The design can also check for 'leakage', that the treatment is effective in one area only and not in other areas. This can be important when the treatment can have detrimental or unknown effects elsewhere.

In Pattern Matching Non-equivalent Dependent Variables design, multiple variables are measured to determine the detail of what is affected and what is not affected.