Why didn't you square the epsilon*a_i term in line 5 of your proof?
Given the prediction error (denoted en+1) can be written as:
Prove it's variance is equal to this (note E(en+1)=0):
At the moment I have this:
But when I sub in for di I can get the first two terms i.e. 1+1/n but I cannot seem to finish the question off, if anyone can help me in any way I would be most grateful.