In case of Principal Component analysis why do we set I2* as equal to d2 equal to I2 - (gamma 1 + gamma 2* I1*)? What is the logic behind setting the industry sector index equal to the error term?
The error term is not correlated to any other term. It is orthogonal which is what we wanted in the first place \[cov(d_2, I_1) = 0\]