Spectral analysis of heart rate variability with the autoregressive method: What model order to choose?

Abstract

This work assessed the influence of the autoregressive model order (ARMO) on the spectral analysis of the heart rate variability (HRV). A sample of 68 R–R series obtained from digital ECG records of young healthy adults in the supine position was used. Normalized spectral indexes for each ARMO were compared by Friedman test followed by the Dunn's procedure and statistical significance was set at P<0.05. The results showed that the AR method using orders from 9 to 25 produces normalized spectral parameters statistically similar and, hence, the algorithms commonly employed to estimate optimum order are not mandatory in this case.