How to do Experiments Flashcards
Select variables
IVs (manipulate)
-Values are chosen and set by experimenter
• Multiple levels
-To test predictions about causation
DVs (measure)
- Dependent on subject’s response
- Always must be a quantitative measure on some level
Determine factors to be kept under control
- Hold constant all other factors that could influence subject’s behavior
- If factor would not affect all levels of IV the same, then must be controlled
Determine random variables
-Factors NOT controlled, allowed to vary randomly (equal probability of occurring, minimizing effects)
-Commonly randomize
• How subjects are assigned to levels of IV
• Order of trials within experiment
-Randomization with constrains
• Experimenter restricts complete randomization in some way
Eliminate potential confounds
-Confounds
• Vary as the IV varies and can affect the DV
• If present, cannot draw any conclusions
• Solution: make it a control variable
-Flaws that are not confounds
• Inadequate # of subjects, responses not properly recorded, etc.
Experimental design
- Select variables
- Determine factors to be kept under control
- Determine random variables
- Eliminate potential confounds
Internal validity
- Better for basic research
- How likely is it that IV CAUSED change in DV?
- Greater exp. control -> more INTERNAL validity
Threats to internal validity
- can cause changes in behavior, other than IV- potential confounds
1. History
2. Maturation
3. Testing
4. Statistical regression
5. Selection
6. Mortality
7. Interaction of selection with another threat
History
- Events in outside world can cause behavioral change
- Must assess likelihood what this could affect behavior measured in study
-Solution: study as short as possible, isolate from outside events during study
Maturation
- People change over time, independent of what happens in study (age, experience, health)
- Solution: short study, include no-treatment control group
Testing
- Act of taking test can change behavior, regardless of treatment
- Solution: only test participants once OR give extensive practice prior to first test OR compare to control group with no treatment
Statistical regression
- Change in test scores due to “regression to the mean”
- Solution: don’t chose participants based on extreme scores (b/c more likely to regress regardless of IV)
Selection
- Groups differ BEFORE IV administered (possible when no random assignment)
- Solution: match groups on relevant variable(s) OR random assignment if possible
Mortality
- Participants may drop out of study before completion
- Solution: use rewards, use placebo so mortality is equal overall
Interaction of selection with another threat
Selection X history
Selection X maturity
-Group differences that emerge AFTER study begins, due to threat affecting one group more
-Solution: match on all relevant variables, minimize threats that tend to interact with selection
External validity
- Better for applied research
- How generalizable are the results?
- Greater exp. control -> less EXTERNAL validity
Approach when validity threatened
Several well controlled studies, each differs somewhat in subject, characteristics, procedures, etc.
