Not all combinations (X,Y) will work of course. If you do not need a theoretical answer, and are content with something that might work, consider generating a sample any way you want, then trying to adjust it to fit your constraints:
>> X = 5;
>> Y = 7;
>> z = randn(100,1);
>> a = fminsearch(@(a) (X-mean(a(1)+a(2)*z))^2 + (Y-mean(abs(a(1)+a(2)*z)))^2,[5 5])
a =
5.3484 7.4122
>> mean(a(1)+a(2)*z)
ans =
5.0000
>> mean(abs(a(1)+a(2)*z))
ans =
7.0000
