Multi-Level Experiments · How to explore differences when you have two independent variables · Identifying variables and levels · Spotting interactions · Drawing interactions Theorising Hypotheses - deduction - predictions - observation - test of predictions - induction Identifying the Variable A forensic psychologist investigates how psychopathy and the type of crime committed affects the way convicted criminal's express emotion: IV - psychopathy, crime committed (self-selecting participant variables) high and low groups DV - emotion IV's must be orthogonal (statistically independent). If they are related, they are oblique (purposefully misleading) Comparing Four Groups Group 1a (HPfraud) Group 1b (HPmurderer) Group 1 (HPsycho) Group 2a (LPfraud) Group 2b (LPmurderer) Group 2 (LPsycho) H1 - High psychopathy < Low psychopathy H2 - Fraud < Murder H3 - Low psychopathy murderers < all others Main Effects Group A (Fraudster) Group B (Murderer) · Comparing the high psychopathy participants with low ones is a main effect of psychopathy . Comparing the fraudsters with murderers is a main effect of type of crime . Main effects with two levels are like t-tests . Main effects with three or more levels are like one-way ANOVAS Interactions 0 High Psychopathy Low Psychopathy 12 10 8 Mean Emotional Expression 6 4 2 We say the effect of psychopathy on emotional expression is moderated Fraudsters Murderers
by type of crime. Or that the effect of type of crime on emotional expression is moderated by psychopathy ANOVA (analysis of variance) tries to partitions the variance in the DV Stage 1 Total variability Between treatments Within treatments This is error (residual) Stage 2 Factor A variability Factor B variability Interaction AxB variability Theory of ANOVA Total variance (SSt) = Variance explained by model (SSm) due to manipulation or Unexplained variance (SSr) due to unknown factors . Compare the explained variance with the unexplained variance (F ratio) . If the model explains more variance that it doesn't then it is considered a good fit (SSm with be greater than SSr), larger the F ratio . F ratios tell us that there is an effect overall SSPS produces an ANOVA table, produces 3 F ratios for a 2x2 ANOVA: (1 for main effect of IV1 and one for main effect of IV2 and one for the interaction effect) Each much be reported independently using the appropriate marginal means tables