the scientific method observation theory hypothesis test making it public peer review replication
TRANSCRIPT
The Scientific Method
• Observation
• Theory
• Hypothesis
• Test
• Making it public• Peer review• Replication
Experimental Research
• Why do experiments?• The problem of causation
Showing Causality
• Experiment• Experimental group and control group• Random assignment to groups• Independent variable• Dependent variable• Differences?
Variables
• Independent• What you think CAUSES the Dependent
Variable to change• Independent bc RESEARCHER controls it
• Dependent• What you think will BE CHANGED by
changes in the Independent Variable• Dependent bc IV controls it
Variables
Arrow of causality
IV DV
Another Variable
• Confounding variables• The ones that confuse things• The ones that MIGHT interfere in the
causal relationship between IV and DV, OR might just look like they do
• Also known as intervening variables
Experimental and Control Groups
• The experimental group gets the level of the INDEPENDENT VARIABLE that you think will CAUSE a CHANGE
• The control group gets the USUAL level of the IV, where you would expect NO CHANGE
Experimental and Control Groups
• Examples• EG gets new medication, CG gets placebo• EG gets new medication, CG gets standard
current medication• EG gets new therapy, CG gets old therapy• EG gets new therapy, CG gets wait list• EG gets new sales pitch, CG gets old one• EG gets new interview, CG gets old one
EG and CG, cont.
• How do you know it’s the level of the IV that made the difference?• Try to make EG and CG as similar as
possible.• Random assignment.• Assignment with matching.• (what are the sources of confounding
effects in these designs?? Interaction between the matched pairs).
Sampling
• We can’t measure or test everybody.• Take a sample.• A RANDOM sample is more likely to
accurately reflect the population, and less likely to bias the results.
• A bigger sample is better … to a point.• Our sample may not be an accurate
representation of the population anyway.• Stats can help.
Statistics
• Results from control group
• Results from experimental group
• How to know if the difference is chance or a “real” difference?
• Math can help.
Researcher Bias
• Hypothesis – we all like to be right (and some of us get paid to be right).
• Double-blind
Quasi-Experimental Designs
• When you cannot randomly assign participants to EG and CG, but you CAN control the IV.• Ex. Do boys respond better than girls to a
new strategy for teaching math?• (4 groups; B CG, B EG, G CG, G EG)
• Can suggest a causal relationship.
Quasi-Experimental Designs
• Ex-post-facto designs: you cannot control the IV • Ex; do smokers or non-smokers have more
fires in their homes?
CANNOT ESTABLISH A CAUSAL RELATIONSHIP
For next class
• Bring a MAGAZINE of ANY kind
Designing an Experiment
• Group of two - Exercise; 3.3%
An Exercise on Variables
• Group of 3 or 4 people
• Likely to work together all semester
• Know each other well enough to make groups on your own, or shall I assign?