Week 5 - Experimental Methods Flashcards
(14 cards)
Why do we need experimental methods?
To demonstrate cause and effect (which observational/ correlational research cannot do - directionality, third variables)
Experimental research
Manipulate IV only, measure DV, compare scores between conditions
Threats to internal validity
Internal validity - trustworthy causal relationship between IV and DV
Extraneous variables - other variables that potentially have an effect (these must be controlled)
Confounding variables - other variables that DO affect
Advantages and disadvantages of experimental methods
Advantages - can establish causality
Disadvantages - need to know what to manipulate, what to measure, need precise hypothesis, need strict control
Types of experimental designs
Between-subjects - participants provide data for one condition only
Within-subjects - participants provide data for each condition
Between-subjects designs
Comparisons made between groups, gives independent scores
Analysis - independent t-test or one-way ANOVA (Mann-Whitney, Kruskall-Wallis if non-parametric)
Advantages - no order effects, no time-related factors
Disadvantages - requires more Ns, individual differences, environmental variables
Individual differences
Differences between people that may become confounds (selection bias)
Can create high variance
IDs can become accidentally confounding (via experimenter or participants)
Avoiding assignment/selection bias - aim for random assignment, hold participant variable constant (external validity danger), restrict participant variable range
Restricted randomisation (pseudorandomisation) - create equivalent groups based on certain variable, pair off to match for confound and randomise within pairs
Individual differences and statistical variance
Minimising individual differences is beneficial as it lowers within group variance and increases between group variance
Environmental threats
Testing at different times or places (easily avoidable - make conditions similar)
Other between-subjects threats
Differential attrition
Diffusion (talking between conditions)
Compensatory equalisation (bringing conditions closer out of pity)
Compensatory rivalry (one condition working harder given condition)
Resentful demoralisation (one condition giving up)
Within-subjects designs
Participants provide data for each condition
Analysis - paired t-test or repeat-measures ANOVA (Wilcoxon’s or Friedman’s ANOVA for non-parametric)
Advantages - no individual difference threats, no assignment bias, fewer Ns needed, more powerful
Disadvantages - environmental threats, time-related factors, order effects
Time-related factors
History effects (confounding external events)
Maturation
Regression to the mean
Instrumentation (altered measurement instruments)
Order effects
Carryover effects (learning carried over to time 2)
Progressive error - practice, fatigue
Solutions - choose between-subjects, control time, counterbalance conditions
Counterbalancing
Varying order of treatment
Can be tricky with multiple conditions (can use Latin square though to start with)
