post test with randomized treatment and control group:
O1 T O2
- Very strong, robust design.
- Requires a control group which is comparable to the treatment group.
- Problem: what does 'comparable' mean?
treatment and nontreatment (control) —> External, confounding effects can be considered random!! (single vs. double blind).
- Randomization avoids selection threat (results biased because participation in treatment not random).
- Often not realizable for ethical or practical reasons.
- Example: COB Scantron vs. Akindi experiment.