Pretest-posttest designs have probably been the most popular approach to research since Man first noticed he could make changes in his environment - and other Men (please note that I include females when I say Man - the limitations of the English language . . . . )

It simply makes intuitive sense to measure something, do something, and then measure again to see what sort of change occurred. It also makes intuitive sense to subtract Time 1 from Time 2 measures (or vice versa) and then determine whether the gain (or loss) is statistically significant. . . .

. . . . until 1975.

This was when J.C. Nunnally published an article that pretty well slammed the use of * gain scores* in t-tests.

Since Nunnally's article, some controversy has arisen. Not everyone agrees with him. Furthermore, there is a t-test technique which many regard as quite acceptable when comparing gain (or loss) scores of a control and an experimental group.

I will be adding this and other computational formulae to this tutorial later on. For the time being, an excellent reference for this type of t-test (as well as others), including a computational worksheet, can be found in:

Fitz-Gibbon, C. T. and Morris, L. L. (1987).How to Analyze Data.Newbury Park, CA: Sage.

Another useful reference, with a slightly different point of view:

Bruning, J.L. and Kintz, B.L. (1968).Computational Handbook of Statistics.Glenview, IL: Scott, Foresman and Company.

Here is an example of a Pretest-Postest application in a medical context.

Return to the discussion of experimental designs.