Analysis of covariance

1
Analysis of covariance
2
When

• ANCOVA is an extra on an ANOVA
• ANOVA
• outcome number of words learned
• IV Sex
• ANCOVA adds a covariate
• covariate size of first language vocabulary

3
Control

• Variables that influence the outcome
• Manipulate
• Control
• hold steady
• randomise
• enter into the statistical model

4
Control

• Hold steady
• Everyone has same L1 vocab
• Mean of groups is same for L1 vocab
• Randomise
• sometimes you cant (e.g. Sex)
• ANCOVA record L1 vocab and allow statistically

5
ANCOVA in SPSS

• Use GLM / Univariate
• Enter outcome as dependent variable
• Enter Sex as Fixed IV
• Enter L1 vocab as covariate

6
ANCOVA assumptions

• Linearity
• relation between covariate and outcome is
  (roughly) linear
• Homogeneity of regression
• relation between covariate and outcome is the
  same for both Men and Women
• relation is the same at each level of the IV

7
Checking Homogeneity of regression in SPSS

• A second run through Analyse/GLM/univariate
• Click the model button
• Select IV (Sex) and Covariate and click across
• Now select both together and click across
• Continue as before.
• Interaction effect must not be significant

8
MANOVA

• Multiple dependent variables
• converted into one composite DV
• Assumptions see Pallant SPSS Survival Guide
• (similar to Multiple Regression Linearity,
  Mahalanobis distances) Plus homescedasticity (use
  Boxs M with a conservative p-value)
• Further reading Tabachnik Fidell