Lecture 7 - Within Subjects Anova Flashcards
What is a within subjects design?
It is the same thing as a repeated measures design
It involves comparing mean scores between the participants, each of which takes part in all IV conditions
What are two advantages to within subjects/repeated measures design?
controls for individual differences
fewer participants are required in total as each takes part in all IV conditions
What are five disadvantages to within subjects/repeated measures design?
Practice effects
Boredom effects
Differential carry-over effects
Sometimes the data isn’t completely independent (and this is an assumption of ANOVA)
It isn’t always possible for all designs - e.g. if the IV is gender you cant really take part in both conditions
What are differential carry-over effects?
the treatment condition affects participants’ performance in a later condition in one way, and in another way when followed by a different condition
How many f-ratios do we calculate?
two
one for the main effect
one for the subject variables
How many components can we split the deviations into?
two
1 - between treatments component (measures effect plus error)
2 - within treatment component (measures error alone)
How do we interpret a Tukey HSD test?
It involves calculating a critical difference and any mean difference (called q in the table) greater than this critical difference value is significant
However, we may just be given stars next to the table which represent the level of significance
Then we go on to describe the results - e.g. ‘A post-hoc Tukey HSD test (p
How do we write a table of means and standard errors?
This should be the FIRST thing you do in your interpretation section for ANY kind of ANOVA
Make a table with the means and standard errors written like so ‘20(1.581)’ (the standard error is in brackets
Above this should be the condition name
Label this table something like ‘Table 1’ and underneath write ‘Table 1 (above) shows the means (and standard errors) for ……………………………………….’
What happens if you find a significant main effect of the subject variable
It is very common and not usually a problem
It is only a problem if…..
1 - specific predictions have been made in advance about performance
2 - there is a hidden aptitude treatment interaction (e.g. skill at crosswords pre-test wasn’t taken into account as they only looked at the effect of practice)
How do we construct an error term in a two way repeated measures design?
There are two options
1 - we may construct an overall error term for all effects (but we may well overestimate the error for individual effects, so this isn’t very good)
2 - we may construct an error term for each area of interest (this is good as we are unlikely to overestimate the error for individual areas of interest)
In a within subjects/repeated measures design, each effect has its OWN error term
what are the assumptions of a within subjects/repeated measures ANOVA?
Same as all ANOVA: 1 - continuous level data 2 - normal distribution 3 - independence of scores 4 - homogeneity of variance
Specific to within subjects/repeated measures
1 - Sphericity/compound symmetry (this means homogeneity of different treatment variances.)
How and when do we test for sphericity
We don’t test for sphericity if each IV only has one or two levels
SPSS does ‘Mauchly’s test of sphericity’
If it is significant = we can assume homogeneity of variances
If it isn’t significant = we cannot
Why do we test for sphercity?
So that we can correct any violations or do an alternate test - if we do not then we are at an increased risk of making type 1 errors
What alternative tests can SPSS do when sphericity has not been met?
Greenhouse-Geisser (liberal so an increased chance of type 1 errors)
Huynh-Feldt (conservative so an increased chance of type 2 errors)
