The aim of the present study was to evaluate different analysis methods for revealing heart rate variability (HRV) differences between untreated patients with Parkinson\'s disease and healthy controls. HRV in standard cardiovascular reflex tests and during a 10 min rest period were measured by time- and frequency-domain and geometrical and non-linear analysis methods. Both frequency- and time-domain measures revealed abnormal HRV in the patients, whereas non-linear and geometrical measures did not. The absolute high-frequency spectral power of HRV was the strongest independent predictor to separate the patients from the controls (p = 0.001), when the main time-domain and absolute frequency-domain measures were compared with each other. When the corresponding normalised spectral units, instead of the absolute units, were used in the comparison, the two best single measures for separating the groups were the 30/15 ratio of the tilting test (p = 0.003) and the max/min ratio during deep breathing (p = 0.024). When the correlations between the different measures were estimated, the time-domain measures, fractal dimension and absolute spectral powers correlated with each other. The frequency- and time-domain analysis techniques of stationary short-term HRV recordings revealed significant differences in cardiovascular regulation between untreated patients with Parkinson\'s disease and the controls. This confirms cardiovascular regulation failure before treatment in the early stages of Parkinson\'s disease. The HRV spectral powers, in absolute units, were the most effective single parameters in segregating the two groups, emphasising the role of spectral analysis in the evaluation of HRV in Parkinson\'s disease.