Week 7-Complex Within Subjects ANOVA ON EXAM Flashcards
What is Complex within subjects ANOVA?
Complex ANOVA designs involve more than one independent variable
We are predicting an interaction effect – the effect of one IV (subject topic) on the DV (well-being ratings) IS DEPENDENT on the other IV (whether the student previously look A-level statistics). The IVs interact with each other.
What do we get from a Complex design?
Two main effects:
–One for each independent variable – here, we will get a main effect of subject type and a main effect of familiarity.
-Each main effect tells you whether there is a difference between conditions, separately for each IV.
-For example, the main effect of subject type tells us whether there is an overall effect of subject type on well-being – this is the same as what we did in lecture 5.
-We also have the main effect of familiarity with statistics
An interaction:
This tells us whether the effect of one IV on the DV, is dependent on the other IV
What are the main effects?
-One for each independent variable –
-These can be viewed almost like a one-way ANOVA in their own right
-You treat the just like you would a one-way ANOVA
-If significant they need post hoc testing UNLESS THE IV ONLY HAS TWO LEVELS
-You wouldn’t post hoc the main effect as you only have 2 levels because it will tell you the same thing twice (ANOVA will tell you if there is an interaction or a significant difference; you only need a post-hoc test to see and explain the differences in 3 or more levels)
What is an Interaction?
-This is the new bit today- a significant interaction just tells you that the IVs work to influence the DV in some way
-(IF SIGNIFICANT) always need post hoc testing.
What is an issue with the design of a within-subjects ANOVA?
We test one participant loads of times
What critical considerations must be taken into account when designing experiments?
-Carry-over effects
-Practise effects
-Fatigue effects
-Awareness
What are Carry-Over Effects?
-This is where one condition “bleeds” into another
-For example, imagine we got our participants to do the high stress, medium stress and control conditions on the same day.
-A participants who does high stress followed by control, is likely to be still stressed when the do the control condition (even with a few hrs gap)
-Someone who does control then stress will have no carry over.
-Doesn’t have to be a something obvious like a major stressor, pharmacological intervention (drug trials just need a “wash-out” period etc.,)
What are Practise effects?
-People get better at tasks, this is particularly problematic for cognitive tasks
-An originally effortful task can change into an unconscious task underpinned by procedural memory
-People may develop strategies to improve performance- e.g. defocusing on a colour conflict Stroop
-Some tasks are based on how quickly a rule is learnt, for example the Iowa gambling task, most people learn that low risk low pay off decks are better than high risk high pay off decks
Improvements:
-Practice trials, buffer trials
-Practice sessions
-Reduce number of trials
How can Practise effects be minimised?
-Use the minimum number of trials in a task that will also give a reliable result
-If using multiple sessions, try and do the minimum number
-Practice trials or sessions
-Large gaps between tests people will forget task parameters, e.g. all sessions a least a week apart.
-Ensure your tasks are suitable for within subjects measures
What are Fatigue Effects?
-Performance can decline due to tiredness and boredom
-Often an issue in cognitive tasks that require a large number of trials
-Look for the minimum, but reliable, number of trials you need on a cognitive task; often this will be the most common variation of that experimental task
-Response acquiescence (not just in within-subjects designs) in anything involve a lot of testing (e.g. big questionnaire-based studies)
-Just responding the same way to all trials
How can Fatigue Effects be reduced?
Several things to reduce this:
1. Breaks in the task, allowing participants to rest
2. Put in “catch trials” where a message on the screen says press button “X” instead of usual responses
Data cleaning measures (like pruning synapses):
-For example, in a simple reaction time based task I would exclude all trials
that have reaction times less than 200ms (pre-emptive responding) and above 2000ms (loss of concentration).
-Where the individuals RT is 3 SDs above their own mean – as this would be an outlier in an individual’s behaviour
What are Latin Squares?
-In an experiment with four conditions there are 24 orders (1x2x3x4)
-Therefore Latin squares are used as there are too many orders to effectively counterbalance
-Each condition occurs once in each position
-Allows us to effectively counterbalance as each condition is in one position
What are the 2 key advantages of Within-Subjects Design?
- Control for individual differences in participants
- Need less participants (increased power)
What is the fundamental problem of testing people several times?
It can improve performance or cause a decline in performance
What are the assumptions of Within Subjects ANOVA?
-Normally(ish) distributed data
-Ratio, interval (or ordinal data)
-Sphericity
-Similar to homogeneity of variance in the between subjects ANOVA
-Sphericity = equality in the differences in variances between levels.
-ANOVAs are very robust (if data isn’t fully normally distributed its ok)
