Lectures (Midterm II) Flashcards
Factorial designs
More than one IV measured at a time. All combinations of IVs represented, so 2x2, 2x3, etc. Use factorial instead of looking at variables one by one because there may be interactions between the IVs - effect of one independent variable may depend on level of another variable
Main effect
In a factorial design, not an interaction. The effect of a change in one IV independent of other variables
Identifying Interactions in Line and Bar graphs
Line graphs - lines not parallel. Bar graphs - difference between bars is unequal
3 things to look at in a 2x2
(1) Main Effect I
(2) Main Effect II
(3) Interaction
Correlation
Measures strength of relationship between variables
Pearson’s r
Measures linear correlation
Ways of comparing means
t-tests for two groups. Can be independent samples t-test (between subjects) or paired samples t-test (within subjects).
ANOVA - more than 2 groups or if there are interactions of more than 1 IV
Type I vs. Type II Error
Type 1 = False Positive, incorrectly rejecting the null hypothesis.
Conditions that make it easier to find differences between groups/find significant results
Smaller variability ->smaller overlap in distributions ->easier to find differences. Do this by eliminating noise and increasing controls
Big effect size - difference between means
Larger sample size - better representation of true distribution
Possible challenges to internal validity
Confounding variables, history, maturation, instrumentation (problems with consistent methodology and observer objectivity), statistical regression, biased selection of subjects
Why is a pilot study important?
Manipulation check - whether manipulation of independent variable really accomplished anything. Feedback. An idea of what results might be like.
Correlational studies
Examines relations among DVs w/o experimental manipulation. If there are no groups, use correlation. Often difficult to establish causality
Two kinds of surveys
Open-ended or categorical
Scale biases
Central tendency - not rating extreme
Acquiescence bias - always agreeing with item
Social desirability bias - say things that have good social meaning
Non-experimental designs
Case studies, single-variable (does behavior differ from chance for ex.), multiple-variable
Experimental designs
Between subjects (randomized or matched-groups), within subjects, mixed design
Error Variance
Variability may not be caused by the IV if there are individual differences between subjects, inconstant environmental conditions, fluctuations in the physical or mental state of participants.
Within-subject advantages and disadvantages
Less error variance from individual differences, making it easier to detect differences. However, fatigue or carryover effects
Mixed design
Partially within subjects and partially between subjects
