We propose a new model for RR interval records based on the empirical finding that the amplitude of RR records is negatively correlated with its regularity. We develop the mathematics needed to define rigorously this model, as well as the statistical tools necessary for estimating the parameters of the model on sampled data. We use this to analyze finely the impact of ventricular extrasystole of RR records, and to assess the efficiency of certain treatments on this pathology.