III. EXPERIMENTAL METHODS

Their goal is establishing cause-effect relationships between independent and dependent variables by eliminating alternative hypotheses; uses quantitative data. The underlying model is that the world works in a linear, cause-and-effect way and that science should identify causal laws.

A. QUASI-EXPERIMENTAL DESIGNS: able to eliminate some but not all alternative hypotheses. Examples include static groups, non-equivalent groups, and time series designs. Used most often in program evaluation, evaluation of public policies, and other research where complete control of the experimental setting is not possible.

B. TRUE EXPERIMENTAL DESIGNS: If well done, these designs eliminate many alternative hypotheses. Include both field experiments and laboratory experiments. Two main types: between-groups or randomized groups designs where each subject gets only one treatment and repeated-measures designs where each subject get more than one (or all) treatments. These two types may be combined in complex designs.

C. SINGLE-SUBJECT DESIGNS: experimental designs with one subject; variation on repeated-measures designs; includes baseline, reversal, changing baseline

D. META-ANALYSIS: statistical re-analysis of previously analyzed experimental data to compute treatment effects