.png){kind=icon}
Loss of subjects varies across experimental conditions.
This design eliminates the possibility of testing-X interaction.
May be confounded with the independent variable when a single group is studied over an extended period of time.
Controls for the effects of history in a trend analysis by comparing a group exposed to the treatment with other groups that are not.
Selection and differential attribution are the primary threats to internal validity of this design.
May occur when measurement of the dependent variable varies over time or experimental conditions.
Resembles the one-group pretest-posttest design except multiple observations are made before and after the treatment.
A potential threat to validity when subjects are selected on the basis of their extreme scores on the dependent variable.
Pretests and posttests are administered to experimental conditions but random assignment is not possible.
The effect of X on Y depends on Z.
A matter of how well an experiment controls for the effects of extraneous variables.
Randomization rules out this threat to internal validity.