High resolution parametric methods offer many distinct advantage over conventional nonparametric methods for numerous applications in smart grid e.g. power quality monitoring and control, islanding detection, low frequency oscillations analysis, etc. Higher computational burden is the main cause of under utilization of these methods, which is dependent on the size of the autocorrelation matrix. The computational burden and the estimation accuracy pose contradictory requirements on the size of the autocorrelation matrix, thus making its selection rather challenging. This paper analyses the effect of various signal attributes (sampling frequency, noise, number of frequency components, number of samples, etc.) on the appropriate dimension of the autocorrelation matrix. The results of this study assist in selecting optimal dimension of the autocorrelation matrix for improved performance of the model based parametric estimation techniques.