ONE GROUP PRETEST - POSTTEST DESIGN

Campbell and Stanley called this a “pre-experimental” design, but I consider it to be experimental (since the X is experimentally manipulated), but with potentially serious problems which we have already discussed:  History, maturation, testing, instrumentation, and possibly regression.  If we have contrived ways to control these threats (which might be possible under if our subjects are inanimate objects whose environments we control completely, as we might imagine things are in the physics or chemistry laboratory), then this design could be OK.  Statistically, the comparison between means on O1 and O2 could be made with correlated samples t or a nonparametric equivalent.

Thanks