This paper studies the impact of stochastic volatility (SV) on optimal investment decisions. We consider three different SV models: an extended Stein/Stein model, the Heston Model and an extended Heston Model with a constant elasticity variance (CEV) process and derive the the long-term optimal investment strategies under each of these processes. Since volatility is not a directly observable quantity, extended Kalman filter techniques are adopted to deal with this partial information problem. Optimal investment strategies based on the CEV volatility model are obtained by adopting the Backward Markov Chain approximation method since analytical solutions are no longer available. We find in the empirical investigation that the Heston model is favored as a more parsimonious model compared with the other two models. All three investment strategies based on the three SV models contain a positive intertemporal hedging term in addition to the static mean-variance portfolio. However, in their details the three investment strategies differ from each other. We also find that the investment strategies are sensitive to the CEV parameter.