Search All of the Math Forum:
Views expressed in these public forums are not endorsed by
Drexel University or The Math Forum.



from anova to ancova
Posted:
Dec 20, 2013 5:52 AM


Hello, I need help. A lot sources indicate that ancova is an anova procedure where dependent vartiables are adjusted(effect of covarites are taken out from dependent variable). I have such a data set; y is dependent variable, x is covariate, and z is the groups. "z" has 3 categories. I transform "z" into dummy variables z1 and z2. Then the regression model is
y=a+b*x+c1*z1+c2*z2+e
By using linear regression, I estimate b; then, I obtain adjusted dependent variable.
yadj=yb_est*x
By using these adjusted dependent variables, I make ANOVA. I also make ANCOVA(Using Univariate GLM procedure in SPSS) to whole data, taking as y as dependent variable, x as covariate and z as fixed effects. However, the sum of squres related to groups are not same. Where does this difference come from? I will be very glad for any helpm. Thanks a lot.



