The analysis of chaotic time series requires proper reconstruction of the state space from the available data in order to successfully estimate invariant properties of the embedded attractor. Using the correlation dimension, we discuss the applicability of the two most common methods of reconstruction, the method of delays (MOD) and the Singular Spectrum Approach (SSA). Contrary to previous discussions, we found that the two methods perform equivalently in practice for noise-free data provided the parameters of the two methods are properly related. In fact, the quality of the reconstruction is in both cases determined by the choice of the time window length w and is independent of the selectedmethod. However, when the data are noisy, we find that SSA outperforms MOD.