Week 8-Mixed ANOVA Flashcards
What is a Mixed ANOVA?
-Mixed ANOVA is just when we use a mix of experimental designs
-The mixed ANOVA is used when we have a within subject IV(s) and between subject IV(s)
-We set up such an experiment if we think a between subjects variable will have a differential effect in/on the different levels of the within subjects IV (and vice versa of course)
-We cover just 2 independent variables not 3 as otherwise just unnecessarily meaningless
How can you know whether an interaction is present via a graph?
If lines converge, diverge or cross you likely have an interaction
What are the disadvantages of Mixed ANOVA?
-Has the disadvantages of both between and within subjects
-In terms of the within subjects factor there may be practice/carry-over/fatigue effects e.g., better at being tactical ‘snake-like’ with practice
-In terms of the between subjects factor the groups may be different from each other in many ways (i.e. variance in individuals behaviour) e.g., Males and females differing in concerns about perception of behaviour
What are the advantages of Mixed ANOVA?
-Has the advantages of both between and within subjects
-Less practice/carry-over/fatigue effects than a “pure” within subjects design
-Less noise from individual differences than a “pure” between subjects design
-Not as many participants as a between subjects design
-Less complicated counterbalancing as a within subjects design
Particularly useful for when your between subjects group already exist (i.e. you cannot randomise people to one condition or to another) like:
-Alcohol-dependent vs. non dependent
-Stroke vs. non-stroke
-Impulsive vs. non-impulsive
-Conservative voters vs. reasonable people
-Males vs. females
What are Quasi-Experimental Designs?
-In a true between subjects experiment we would randomly allocate participants to experimental conditions
-Random allocation should balance participant characteristics between conditions (worth checking!)
-In QUASI-EXPERIMENTS there is no full control over allocation of participants to different conditions (the groups exist outside your experiment).
-This means that participants in each condition may systematically differ on variables other than the dependent variable
What is used to control the confounding variables in a Quasi-Experimental Design?
-In a quasi-experiment it is difficult to ascertain if the between subjects condition is causing an effect or if another factor that is associated with the between subjects condition causing it (confounding variable)
-One way to try and control for this is through a matched design
-In a matched design you ensure that you match your control condition on critical factors that may influence your DV
-For example, imagine we wanted to run a study that looked at the effect of mothers who abused alcohol during pregnancy on attention deficits in children
-In this example we cannot do a true experiment i.e. make some mothers drink heavily during pregnancy and have a control group who do not
What is another way to do quasi-experimental designs?
-Another way of doing quasi-experimental designs is to predict situations in which groups will and will not differ
-This can be simple- for example we may simply want to see if groups differ before and after an intervention.
How can you determine when to test a study using WS or BS?
-You may think one variable may be heavily influenced by between group factors (so deliver this on a within subjects basis)
-You may think another variable may be heavily influenced by repeated testing (so deliver this on a between subjects basis)
What are the assumptions for a Mixed ANOVA?
-Usual ANOVA assumptions
-Normal distribution, interval or ratio data (some ordinal also ok- e.g. attitude scales that are equidistant)
-As we have within and between variables we need to check:
-Homogeneity of variance (Levene’s) for the between subjects factor(s)
-Sphericity (Mauchley’s) for the within subjects factor(s) if 3+ levels!
-We want Levene’s test and Mauchley’s to be insignificant
What is the Equality of covariances matrices?
-Whether the correlations between any two repeatedly measured DVs is similar to the other correlations between all other repeatedly measured DVs
-Box’s test is computed for this.
-You want Box’s test to be non-significant (like other ANOVA assumption tests)
-However, unlike the others it is only an issue if p<.001!
-Issue of p<.001 is specific to Box’s test and NO OTHER TESTS
What is the Summary?
-Mixed ANOVA is an example of a complex design which incorporates both between and within subject variables.
-It has the same advantages/disadvantages to within/between subject designs.
-Consideration should be given when deciding on whether variables can appropriately be between or within subjects
-We check main effects and interaction effect for mixed ANOVAs
-Interpretation of the interaction effect needs to be done with subsequent analyses.
