Please enable JavaScript.
Coggle requires JavaScript to display documents.
RM III - Coggle Diagram
RM III
MANOVA
Assumptions
- Cases independent
- DVs normal
Equivalence of covariate matrices
- Matrices between groups (AB, AC, etc.) approx. equiv.
- Barlett's Test or Box's M; sig -> violation (not good), p < .001
Equality of variance
- Matrices between groups (AA, BB, etc.) approx. equiv.
- Levene's Test; sig -> violation
Use
- More info than multiple ANOVAs
- Uncover relationships between DVs
- Accounts for correlation between DVs -> higher external validity
Variance: how DV scores vary around mean
- Total: individual vs grand = between + within
- Between: group mean vs grand
- Within: individual vs group mean
Df: potential sources of variance
- Between = K - 1
- Within = N = K
Mean SS = SS / DF
F = MS between / MS within (larger = more diff)
- df larger -> critical F value lower
- Doesn't tell which group sig diff
Eta-squared = SS between / SS total
- Equiv to R2
- Larger sample -> more likely sig -> Type I error increase
No. participants
- Determined by effect size
- Requires smaller sample than ANOVA
- Multiple DV -> sensitive to small diff ->
-
Summary
Continuous DV, Cat IV
- t-test: 1 IV with 2 groups
- One-sample: sample vs pop mean
- Independent sample: 2 groups
- Matched samples: 2 timepoints (same people)
- ANOVA: 1+? IV with 2+ groups, 1 DV
- Repeated-measures: 1+ between, 1 within factor
- Factorial ANOVA: 2+ cat IVs
- MANCOVA: 1+? IV with 2+ groups, 2+ DV
Cat DV
- Chi-square: 1 cat IV
- Logistic: 1+ binary IV
Continuous DV, Continuous IV
- Simple linear regression: 1 continuous IV
- Multiple linear regression: 2+ continuous/cat IV
- MLM: 1+ levels of IV
Regression
Multiple
Relationship
- 1 continuous DV
- 2+ continuous/(categorical) IV
RQ
- How much variance explained in DV by IVs
- => R2, Adjusted R2 (for sample size & no. predictors), F-test for sig
- Also R (correlation)
- Which IVs most important?
- => beta-weight, t-test for sig
- Also b-weight (unstandardised, for predicting scores)
- Which IVs explain most unique variance?
- => semi-partial correlation sr2
- Hierarchical: how much additional variance explained after controlling for IV1 and IV2?
- => Change in R2, F-test for sig
Assumptions
- Outliers
- Homoscedasticity
- Singularity / multicollinearity
- Independent errors
- Normality
- Linearity
- Sample size
-
-
-
-
-