Week Eleven - ANCOVA Flashcards
What is an ANCOVA?
A statistical procedure
Normally, what are secondary variables controlled by?
elimination, constancy, or turning them into an IV.
What are we doing with ANCOVA?
We are statistically controlling for the offending secondary variable by removing the variance associated with it.
eg adjusting everybody’s scores on the DV to be what we would think if everyone had the same score on the covariate
What is the major benefit for ANCOVA?
It allows a more sensitive test of treatment effects as we are removing the variance that is due to differences on the covariate which reduces the amount of unexplained variance.
What does an ANCOVA produce?
main effect of treatment
main effect of covariate
interaction between the treatment and CV
What does the main effect of treatment test in an ANCOVA?
Tests whether the means of the groups are sig difference after controlling for the effect of the CV on the DV
- are the intercepts for the regression lines for each group the same? - if not: there is a sig difference between the groups on the DV when the CV is 0
What does the main effect of covariate test in an ANCOVA?
Tests whether the CV is sig related to the DV - if not this is good cause CV not going to have impact ov DV scores.
- within the treatment groups, is there a correlation between the scores on the DV and the CV.
What does the interaction between treatment (IV levels) and the CV test in an ANCOVA?
Tests whether the relationship between the CV and DV is the same for all treatment groups
A sig interaction would tell us whether the slope of the DV-CV regression line is the same for each treatment - if sig its not good (size of differences between groups will change)
What is an important assumption of ANCOVA?
Parallel relationships between the covariate and the DV for both groups
What happens if the slopes/regression lines are not parallel (non homo) what happens?
The size of the difference between groups is not going to be the same at the levels of the covariate
What does ANCOVA not include?
Does not include an interaction term as it assumes the slope for all groups are the same
When doing the analysis, when do you know that an ANCOVA is not the right analysis to be doing?
When you don’t have a significant main effect for the covariate
What does an ANCOHET do?
It explicitly tests whether the slopes of the regression line between DV and CV are homogenous
Assumptions of ANCOVA and ANCOHET?
All normal ANOVA ones plus
- linear relationships between CV and DV
- normal distribution of scores
- HOV
- independence of residuals
- homo of regression: parallel slopes
What is the difference between ANCOVA and ANCOHET in regard to homogeneity of regression?
ANCOVA assumes it
ACOHET tests it