what is an experiment?
they:
1. manipulate an IV and measure a DV
2. Allow causal inference
3. Differ from surveys/interviews/observations which are correlational
what are the criteria for a true experiment?
- manipulation of at least one IV
- Random assignment
- Control over ext. variables
- measurement of a DV
What is triangulation?
Using multiple methods, measures, or data sources to study the same construct (increases validity and reduces bias)
what are some strategies to increase external validity?
- Use diverse samples
- Recreate realistic settings
- Replicate studies
what are some strategies to increase ecological validity?
- naturalistic tasks
- realistic stimuli
- field experiments
what are some strategies to increase internal validity?
- Random assignment
- Control groups
- Remove confounds
- Standardized procedures
what is experimental control?
Holding everything constant except the IV
what is statistical control?
Using models to adjust for covariates
clarify single-item vs index measurements
- Single-item indicator: only one question used for very concrete concepts (i.e: age)
- Index: more complicated concepts measured with more than one question (ex: attitudes, feelings, judgements, norms, culture, complex identities)
what is face validity?
when it looks like the measure accurately represents the construct
what is discriminant validity?
when the construct is not correlated with unrelated constructs
what is construct validity?
how well the study’s measures and manipulations reflect theoretical constructs
what is convergent validity?
when the construct does correlate with similar constructs
what is criterion validity?
evaluates how accurately a test measures the outcome it was designed to measure.
what is a factor?
a manipulated variable
what is a within-subjects design
a design where all subjects are exposed to all conditions
con: carry-over effects are likely
what is a between subjects design?
each subject is exposed to only one condition
gold standard
what are main effects?
when levels in one factor alone cause changes in outcome variable; difference is in levels of one factor with nothing to do with other factors
what are interaction effects?
when the effect of one factor changes dependent on the outcome of another factor
what experimental power?
the probability of acheiveing statistical significance when an effect actually exists
OR
the probability of supporting a hypothesis (or rejecting the null)
how do we increase experimental power?
- Larger sample size
- Strong manipulations
- Reliable measures
what is statistical significance?
whether the effect is unlikely due to chance (p < .05)
how do we create excessive study power?
excessively large sample
what does a effect sizes tell us?
it’s a practical measure of the significance of the relationship between the IV and DV
complementary to p-values
