Experimental and Quasi- experimental Designs Having fun with X’s and O’s
Inferring Causality The essence of evaluation Is the “effect” the result of the “cause”? Reality rarely accommodates us The mischief of intervening variables – threats to internal validity The year the Twins won the pennant Other threats to internal validity: p. 317, 328
Experimental Designs The language of experimental design X = intervention O = Observation O 1 = Sequence of observation R = Randomization E = Experimental Group; C = Control The Gold Standard: R EO 1 XO 2 R CO 1 O 2
How a Classic Experiment Works (p. 324) Experimental GroupControl Group Compare: Same? The Intervention Remeasure Dependent Variable Measure Dependent Variable Remeasure Dependent Variable Measure Dependent Variable Compare: Different?
Distinguishing Characteristic of Experimental Designs Randomization: comparability of groups What is the ethical problem? Possible solution: matching Similar to stratification of a sample
Pre-Experimental Designs The Post-test only, or one-shot design XO1XO1 The One-Group Pretest-Posttest Design O1XO2O1XO2 Posttest only with nonequivalent groups (p. 322) XO1O2XO1O2
Quasi-experimental Designs Classic Experimental Design, without Randomization EO1XO2CO1O2EO1XO2CO1O2
Beefing Up Quasi-Experiments Use Multiple Pretests O1O2O3XO4O5O6O1O2O3XO4O5O6
Beefing Up Quasi-Experiments Vary the timing of the intervention EO1XO2O3CO1O2XO3EO1XO2O3CO1O2XO3
Beefing Up Quasi-Experiments Use non-equivalent interventions EO1X1O2CO1X2O2EO1X1O2CO1X2O2
Beefing Up Quasi-Experiments Select a Control group that matches the experimentals on variables of interest: E MO 1 XO 2 C MO 1 O 2