In this study, the effects of methods and techniques of frequency domain measures in the analysis of heart rate variability (HRV) that are used in discriminating the patients with congestive heart failure (CHF) from normal subjects are investigated. Frequency domain HRV measures are obtained from 29 CHF patients and 54 normals. These measures are calculated using wavelet entropy in addition to standard FFT and LOMB algorithms based periodograms. Euclidean and City distance metrics are used as distance and obtained measures are investigated by k-means clustering using three different normalization techniques. As a result, the subset of the frequency domain HRV measures are high frequency components obtained from LOMB periodogram. Using the optimal subset of HRV measures gives discrimination accuracy of 81.93%.