Researchers developed a new method to estimate stochastic volatility models accurately. They compared two likelihood procedures using simulations and found that an approximation method is almost as accurate as the full likelihood method. Applying this method to market data, they found evidence supporting a general CEV model for volatility.