Factor model estimation by using the alpha-EM algorithm

In this paper, we apply the alpha-EM algorithm to factor model estimation. The alpha-EM includes the traditional log-EM as a special case. It has been shown that the convergence speed of the alpha-EM algorithm is much faster than log-EM algorithm. The alpha-EM also allows us to choose different alphas to achieve the fastest convergence speed and more accurate factor model estimation for different problems. In practice the update equations from the alpha-EM are not tractable. We can apply causal approximation and series expansion to those update equations to get practital update equations. With these update equations we can show that the alpha-EM can save us in total computation time. Empirical results from real.nancial data are given.