POSTTEST ONLY CONTROL GROUP DESIGN

Here we simply assign subjects to groups in a way that should assure pretreatment equivalence, don’t bother with a pretest, administer the treatment to the one group, and then measure the criterion variable.  With respect to controlling the previously discussed threats to internal and external validity, this design is the strongest of all I have presented so far.  However, this design usually is less powerful than designs that include a pretest-posttest comparison.  That is, compared to designs that employ within-subjects comparisons, this design has a higher probability of a Type II error, failing to detect the effect of the treatment variable (failing to reject the null hypothesis of no effect) when that variable really does have an effect.  Accordingly, it is appropriate to refer to this threat to internal validity as statistical conclusion validity.  One can increase the statistical power of this design by converting extraneous variables to covariates or additional factors in a factorial ANOVA, as briefly discussed later in this document (and not-so-briefly discussed later in this course).  While it is theoretically possible to make another type of error that would threaten statistical conclusion validity, the Type I error, in which one concludes that the treatment has an effect when in fact it does not (a Type I error), it is my opinion that the Type II error is the error about which we should be more concerned, since it is much more likely to occur than a Type I error, given current conventions associated with conducting statistical analysis.

Thanks