Week 7 & 8 Exam Notes PDF

Week 7 & 8 Week 7: Between-subjects methods Week 8: Within-subjects methods Week 7 2 types of Experimental designs Between-subjects Within-subjects ‘independent measures ‘repeated...

Week 7 & 8 Week 7: Between-subjects methods Week 8: Within-subjects methods Week 7 2 types of Experimental designs Between-subjects Within-subjects ‘independent measures ‘repeated measures’ Different participants in Same participants in each condition each condition  Between-subjects The data contain only 1 score for each participant. Analysis 2 conditions independent samples t-test 3+ conditions One-way ANOVA Advantages Disadvantages Not influenced by time-related factors Requires larger number of participants problem with special populations E.g., history effects, maturation Vulnerable to confounds Not influenced by order effects (threats to internal validity) Practice Individual differences Fatigue Environmental variables  Avoiding selection bias Restricted randomisation Hold variables constant & restrict range Create equal groups E.g. Hold gender constant Have only male participants Match for a potential confound (e.g. age) Restrict range Block randomisation Age Example IQ heads = oldest in group 1 Tails = oldest in group 2 Individual differences and variance Statistical tests assess ratio of between group variance & within group variance Aim for low variance within groups and high variance between groups minimizing individual differences can be advantageous BUT it compromises generalisability (external validity)  Reducing environmental threats to internal validity Run participants at same time of day Use same location Other threats Differential attrition Diffusion (participants) differences in attrition rates spread of the treatment from the from one group to another experimental group to the control group If several withdraw from 1st group, treatment is known in Group 1 and sample size is no longer similar Control Group – reduce the difference to 2nd group between groups Compensatory equalisation (participants) Compensatory rivalry (participants/experimenters) Control group learns about the treatment being received and Control group compensates by working demands the same or equal treatment extra hard to show that they can perform just as well the treatment group Resentful demoralisation (participants/experimenters) give up when they learn about the treatment group become less productive and motivated  Within-subjects repeated measures Advantages Disadvantages Removes or reduces threats from Vulnerable to individual differences Environmental threats No threat from selection bias Time-related factors (it’s the same participants) Order effects avoids increased variance Fewer participants needed (greater statistical power) Analysis 2 conditions Paired-samples t-test (Wilcoxon for non-parametric) 3+ conditions Repeated-measures ANOVA (Friedman’s ANOVA for non-parametric)  Time threats History effects Maturation Time has passed effect is due to increase in age/ development → causing a difference in results Regression to the mean Instrumentation Extreme values (high or low) may Change in coding be less extreme when tested a 2nd time Change in tools/equipment threat to internal validity because (e.g. faulty scales) changes in’ scores can be caused by regression instead of the treatment Avoid this effect by — repeating the test multiple times to rule out luckiness  Order effects Carry over effects Treatment A is influencing the results of Treatment B Progressive error Practice Fatigue reduce order effects: Choose a between-subjects design Control time Counterbalance  Counterbalancing alternating the order of treatments to ‘balance’ time effects aim: eliminate the potential for confounding by disrupting any systematic relationship between the order of treatments and time-related factors. multiple treatment conditions, then you need *N sequences and the number of ppts will be a multiple of *N E.g. 3 conditions = 6 groups E.g. 4 conditions = 24 groups E.g. 5 conditions = 120 groups / different orders Number Factorial method Latin square is one way to counterbalance conditions However B always comes after A, C always comes after B etc so order effects are not completely controlled