Lecture 10 - Multivariate Analysis of Variance Flashcards
When do we use a MANOVA?
When there are two or more dependent variables
How does MANOVA work?
It creates a single new DV from the others. This is a linear combination of the experimental DV’s and it attempts to maximise the differences between treatment groups
What are three advantages of MANOVA?
Improved chance of finding out what changes occur due to the experimental treatment
Since all DVs are combined into one, it only counts as a single comparison and so type one error rate isn’t increased like it would be with multiple comparisons
It has more statistical power than ANOVA so it may show differences that an ANOVA cannot
What are the 5 assumptions of MANOVA
Multi variate normality
Homogeneity of variance-covariance matrices
Linearity
Multicolinearity
Singularity
How can we test for homogeneity of variance-covariance matrices in a MANOVA
We use Box’s M but with a criterion of p
What does multicolinearity mean?
The relationship between pairs of variables is high (r must be greater than 0.9)
What do we do if we find a significant multi variate effect
Then we conduct an ANOVA for each DV so we can look at univariate effects
If we find a significant univariate effect then we conduct post hoc tests where necessary
What are the assumptions of DFA
Same as a MANOVA
What does DFA stand for
Discriminant functions analysis
What does DFA do
It looks for a set of variables that can predict membership of groups
These predictors are normally chosen based on theories
DFA calculates different ‘functions’ that maximises the ability to predict group membership
What is the maximum number of functions in DFA
The number of levels of the grouping variable -1
Or
The number of degrees of freedom of the IV
What is a MANOVA used to work out
Can we predict group membership from a set of predictors? What proportion?
What are the differences between these predictors and how are they associated?
