Analysis of covariance (ANCOVA): model, adjusted vs. observed treatment means, estimated lines, non-parallel lines

4 important questions on Analysis of covariance (ANCOVA): model, adjusted vs. observed treatment means, estimated lines, non-parallel lines

What are two ways to look at analysis of covariance?

-Covariate x is introduced to increase precision of comparison between treatments. We correct for differences amoung the values of the covariate for the experimental units. Technically, we filter out part of the error variation (similar to blocking
-There a specific interest in the relationship between y and x. We want to know whether this relationship is affected by the treatments

What are the extra model assumptions in ANCOVA?

The extra mode assumptions in ANCOVA are:
-there is a linear relationship between response y and covariate x
-slope (coefficient of covariate x) is the same for all treatments
-covariate x does not depend on the treatments

What would happen if x is influenced by treatment?

ANCOVA assums that x is a pre-treatment variable
If treatment affects x, then x contains treatment information ANOCVA removes part of the treatment effect when adjusting for x. You distort the true treatment effect
  • Higher grades + faster learning
  • Never study anything twice
  • 100% sure, 100% understanding
Discover Study Smart

How do you check whether the parallel lines assumption in ANCOVA is valid, and what do you do depending on the result

Fit a full model including a treatment × covariate interaction and test it with an F-test; if the interaction is not significant, assume parallel lines and use the reduced ANCOVA model, but if it is significant, slopes differ and you must use the full interaction model instead.

The question on the page originate from the summary of the following study material:

  • A unique study and practice tool
  • Never study anything twice again
  • Get the grades you hope for
  • 100% sure, 100% understanding
Remember faster, study better. Scientifically proven.
Trustpilot Logo