Quasi-experiments

How is a quasi-experiment different from an experiment?

• Subjects are not randomly assigned to conditions.
• Subjects are selected based on the values of the independent variable, rather than having the experimenter assign values of the independent variable to subjects
• Quasi-experiments have less internal validity than experiments

How are Quasi-experiments different from correlational studies?

• Technically, they are a type of correlational study
• Quasi-experiments usually select only a certain range of values of an independent variable, while a typical correlational study measures all available values of an independent variable.
• Quasi-experiments usually have more control than a typical correlational study (but less than an experiment)

Types of Quasi-experiments

• Quasi-experiments with "subject variables"
• Between-subjects
• examples
• Gender differences
• High vs. low self-esteem
• Quasi-experiments with environmental variables: Time Series Designs
• Within-subjects
• Types:
• One-group pretest-posttest
• No control group
• Very low internal validity
• Avoid if at all possible
• Interrupted Time-Series
• No control group
• Multiple measures of the DV before and after the introduction of the IV
• Example: the effect of the personal computer on productivity
• Repeated-Treatment Design
• No control group
• Repeat the treatment (IV) multiple times
• Obtain pre-test and post-test measures each time
• Example: the effects of Federal Reserve Board rate cuts on economic growth
• Non-equivalent control groups
• Need to have both pre-test and post-test measures of the DV
• An improvement on the one-group pre-test post-test design
• Example: Did the introduction of the death penalty in NY reduce crime? Compare to another non-death-penalty state that did not introduce the death penalty.
• Multiple Time-Series
• Combines the interrupted time-series and non-equivalent control groups designs
• Includes a non-equivalent control group
• Multiple measures of the DV for both groups before the introduction of the treatment
• Multiple measures of the DV for both groups after the introduction of the treatment
• Example: effects of winning basketball championship on gym memberships in Chicago vs. Utah
• Quasi-experiments involving the passage of time: testing developmental changes
• Cross-sectional studies
• Confound: cohort effects
• Longitudinal studies
• Confound: secular trends
• Cross-sequential designs
• Also shows the "time lag effect" - people of a certain age at different times
• Gives a partial estimate of the effects of secular trends

Preventing Smoking in Adolescents name _______________________________

(see textbook, p. 359)

What type of design did you select?

Describe your study design: (IV, DV, control groups if any, when measurements are taken, etc)

What potential confounds, etc. do you see as problems for your design?