ANCOVA with more than 2 covariates. homogeneity of regression slopes.

#1
Hello,

I am carrying out a ANCOVA with more than 2 covariates with SPSS. I want to check the homogeneity of regression slopes to confirm the assumption.

When in a ANCOVA with one covariates, I can understand.
https://www.youtube.com/watch?v=oNX8hL6QHNM

When in a ANCOVA with more than 2 covariates(e.g., 3 covariates), to select factor, covariate1,covariate2,and covariate3, simultaneously in the box labelled 'model' of univariate model window(this looks like factor*covariate1*covariate2*covariate3) is correct?


Thank you.
 

Dragan

Super Moderator
#2
Well, no, that is not correct.

It should be:

Fac
Fac*Cov1
Fac*Cov2
Fac*Cov3
Fac*Cov1*Cov2
Fac*Cov1*Cov3
Fac*Cov2*Cov3
Fac*Cov1*Cov2*Cov3

Get the R^2 for this Full model.

Then get the R^2 for the Reduced model (i.e., without any slope/treatment interaction):

Fac
Cov1
Cov2
Cov3

Once you have these R^2's, compute a F-statistic for the incremental increase in the R^2 for the reduced model and the R^2 for the full model.
 
#3
Dear Dr. Dragan

Thank you very much for your reply.

I could get R^2 for the full model (R^2 =0.178, F=4.590 in 'tests of between-subjects effects') and the R^2 for the reduced model (R^2 =0.147, F=17.674).
(just R^2, not adjusted R^2? Is it OK?)

Would you please tell about how to compute a F-statistic for the increase between the 2(17.674-4.590?), and how to interpret the results?

I really appreciate for your help.