LECTURE 14.3


QUASI-EXPERIMENTAL DESIGNS

I. What the simple experiment accomplishes

A. Cause before effect

B. Observe covariation

C. Randomize extraneous variables

II. What nonequivalent control group design accomplishes

A. Cause before effect (usually)

B. Observe covariation

C. Eliminate some of the extraneous variables, but doesn't control for:

1. Selection

2. Interactions with selection

D. If you match groups,

1. May reduce selection threat, but

2. Still may have interactions with selection and

3. May have regression toward different means

E. Ideally, to maximize internal validity, with this design, you would

1. Match on reliable measure to reduce regression

2. Have a pattern of results inconsistent with regression (control group not changing from pre- to post-)

3. Match on variable that is strongly related to posttest measure (to reduce selection bias)

4. Test subjects immediately after they had been matched to reduce chances of other variables interacting with selection

III. Time-series designs

A. Eliminate selection threats

B. Assume that the effects of history, maturation, mortality, testing, instrumentation, and regression would be consistent or cyclical over time

C. Problem: Effects are not always continuous, especially history effects

D. One solution: Reversal designs

E. Another solution: Combine the strengths of the time-series design with the strengths of the non-equivalent control group design by using a two-group time-series design

F. Examples of two-group time-series designs (drinking age and different states and accidents)

IV. Tactics for Studying Age (Maturation)

A. Cross-sectional designs: A version of the non-equivalent control group design

B. Longitudinal designs: A version of the time-series design

C. Sequential strategies: A version of the multiple-group time-series design

V. Conclusions

A. Quasi-experimental methods are flexible

B. Rely on knowing threats to internal validity and discounting them on the basis of being unlikely rather than using statistical analysis

C. May have impressive external validity


Back to Chapter Menu