I didn't get your question clearly, but as I get, I think you should use cluster. Here is repeated measure ANOVA using PROC GLM:

data;

input a b r y1 y2 y3 ...; (a and b are independent factors, r is repetition, and y's are dependent vars)

y=y1;

y=y2;

y=y3;

....

cards;

.

.

.

proc glm;

class a b r(if needed);

model y1 y2 y3 ...= a|b r(if needed);

repeated y /options;

run;

That will work if all independent variables are general, my independent variable changes for each y, and that is what I don't know how to do.

That is y1 is the willingness to buy a bundle of oranges and apples, and I want to have an independent variable of how much you consume each specific product. y2 will have different products in the bundle, and different data on consumption.

Is there a way to do it with Anova or should I use regression.

So I think you measured several different variables in each subject. In this case you'd better use cluster or discriminant analysis according to your goal.

If the question is Boolean (buy versus not but) discriminate analysis is the way to go.