Module 4 Flashcards
During what part of our ANCOVA analysis do we need to split the data file by groups?
During assumption testing when we’re testing normality/outliers and the independence of the treatment variable and covariate.
What are the two statistical tests of normality? What do we want their results to be?
KS and Shapiro-Wilk. We want them to be non-significant
How do we test the assumption of independence of treatment variable and covariate?
We run an ANOVA between the covariate and the treatment variable. In the output, we go to the ‘Tests of Between Subjects Effects’ table. We want a non-significant result, to show that the covariate does not differ significantly across the groups.
What does Levene’s test tell us? What result do we want?
Levene’s test tells us about homogeneity of variance. We want a non-significant result
In the main ANCOVA analysis, what does the ‘Tests of Between Subjects Effects’ table tell us?
This tells us if our covariate is significantly related to the DV or not. It also tell us if the IV is significantly related to the DV.
What do we want our ‘Observed power’ to be around?
.7/.8
What do we want from our estimated marginal means?
We want them to have increased.
What do we have to check after our main analysis?
We need to check for homogeneity of regression slopes
How do we statistically check for homogeneity of regression slopes?
We have to run the ANCOVA again, this time going into the ‘Model’ settings and adding the two main effects and the interaction.
Where do we look in the output to evaluate homogeneity of regression slopes? What result do we want?
‘Tests of Between Subjects Effects’ table. We want a non-significant result for the interaction, showing that this assumption has not been violated and the analysis is valid.
What are our two different options for reporting effect size? Which is preferred?
We can either report
What are our two different options for reporting effect size? Which is preferred?
We can either report eta squared for each effect or partial eta squared. Partial eta is preferred.
If the partial eta of the independent variable was greater than that of the covariate, what does this tell us?
This means the independent variable explained a greater proportion of the variance not attributable to other variables than the covariate
What type of variance does ANCOVA reduce? What Type (I or II) of error does it reduce?
ANCOVA reduces within-group error variance and the probability of Type I error
What is the 10:1 rule?
The number of covariates you have should be no more than 10% of sample size - (number of groups - 1)
