Editing RR Series and computation of long-term scaling parameters

The editing of heart rate variability (HRV) sequences is largely employed in presence of biological (ectopies, arrhythmias) and technical artifacts. Little is know about the effects of these corrections on the estimation of the long-term scaling exponents, especially for long artifacts. We therefore investigated the robustness of three popular scaling exponent estimators (DFA, a-slope and Dispersional Analysis) with an increasing number of missing RR samples. We tested three editing methods: (i) substitution with local mean value, (ii) linear interpolation and (iii) deletion. Starting from long uncorrupted (> 10000 points) NN series, we artificially inserted artifacts. We then evaluated the effect of the editing methods on the estimation of the scaling exponents. As a reference, the same computation was performed simply removing an equivalent number of points at the extreme of the series. The simulations suggest a negligible effect of the corrections, at least as long as the number of points edited is relatively small.