The paper analyzes the empirical performance between the Stochastic Volatility (SV) and TAR-GARCH models. SV models are flexible enough to explain excess kurtosis, although the inclusion of TAR in the GARCH model improves the variance asymmetry in the time series. These models are used to analyze three daily time series (one simulated series and two financial time series) in order to illustrate the performance of both models. The analysis of residuals is used to evaluate the goodness of fit. We conclude that the SARV model is more useful for capturing the main features of the volatility time series than the TAR-GARCH model.