manova in SPSS Flashcards
Which statistic should you use?
If the IV has 2 levels (df=1), all the different options will have the same F anyway.
If group differences are concentrated on the 1st variate, Roy’s largest root is more powerful (->Hetelling-Lawley Trace-> Wilk’s Lmabda-> Pillai’s Trace)
If group differences are in more than 1 variate Pillai is more powerful (->Wilk-> Hotelling-> Roy)
Which statistic is the most robust?
Pillai’s Trace is generally the most robust most of the time .
Use this is the assumptions have been violated
1)Descriptives box
DESCRIPTIVES BOX:
Looking at the mean scores on the DVs gives an idea of effects
Get a sense of the data
2) Check for homogeneity of covaraince matrices assumption
BOX’S TEST OF EQUALITY OF COVARIANCE MATRICES:
significance (p>.05) means it has been violated
T&F say you don’t need to worry unless it’s p
3) multivariate tests (manova)
MULTIVARIATE TESTS
Choose which F statistic to look at (Pillai’s if all the assumptions haven’t been met).
Ignore the ‘intercept’ section, just look at the ‘condition’ or ‘treatment’ (IV) section.
This tells you if there is a main effect of ‘condition’ (IV) on the LC.
When reporting, include partial eta-squ as a measure of effect size e.g. .49 tells you 49% of variance in the LC is explained by the IV.
4) Homogeneity of variance (2)
LEVENE’S TEST OF QUALITY OF VARAINCES
:) more robust against non-normality than Bartlett’s or Chronbach’s
Significance= the assumption has been violated (needs to be p>.05)
5) Univariate anovas
TEST OF BETWEEN SUBJECTS EFFECTS
SPSS also provides univariate anovas
:( these test don’t consider how DVs may have overlapping variance with each other and/ where some might be more important
6) Host hoc tests
MULTIPLE COMPARISONS
(e.g. Tykey’s HSD)
Used to determine WHERE the group differences are (i.e. which conditions differ)
This does all the possible comparisons between each level of IV on each DV