F Change Test and ANCOVA

#1
Given R squared between the DV and covariant, R squared between the DV and treatment (two groups), R squared between DV and covariant treatment combined, and N, could you conduct an F change test? If so, how? Could you make any conclusions about the effect of the treatment?
 

Dragan

Super Moderator
#2
Given R squared between the DV and covariant, R squared between the DV and treatment (two groups), R squared between DV and covariant treatment combined, and N, could you conduct an F change test? If so, how? Could you make any conclusions about the effect of the treatment?
Yes, you can. Note that ANCOVA is a combination of ANOVA and Regression Analysis. As such, the "F change" statistic becomes nothing more than a t-statistic on the covariate (Square Root of the F-statistic) because it is based on one - degree of freedom. So, the way to handle this is look at the R-square for the full model (that includes the covariate) and then exclude the covariate from the model (ANOVA) and subsequently do the F-Change statistic on 1, and N- 2- 1 degrees of freedom.