What's in a Mean? Comparing Interbeat Interval Averaging Methods Across Variability Levels and Window Lengths

ABSTRACT

Measures indexing the timing of the cardiac cycle, including interbeat intervals (IBIs; also known as heart periods), are some of the most widely used in psychophysiology, and recent guidelines emphasize transparent reporting of how these metrics are calculated. Although duration‐weighted averages—in which each IBI is weighted by its duration within a time interval—have long been recommended for mapping IBIs onto real‐time units, some software programs compute unweighted averages by default, and few published studies report which method was used. This raises an important question: how consequential are disagreements between these two approaches? To address this, we simulated 5 min of continuous IBI data for 10,000 participants under different levels of within‐person variability in consecutive IBIs (with the maximum change ranging from 1% to 25%). We then aggregated the data within nonoverlapping windows of varying lengths (between 250 and 300,000 ms) and compared duration‐weighted and unweighted IBI means. Greater within‐person variability and longer window lengths were each associated with larger average discrepancies between the two methods. Analyses of real, resting IBI data from over 500 healthy adults showed consistent patterns. Longer windows were associated with increased average disagreement, and the magnitude of disagreement was comparable to that observed in the simulation under similarly low levels of within‐person variability. Under typical resting conditions—that is, when within‐person variability in consecutive IBIs was low—average disagreement was minimal. However, at high levels of within‐person variability and at longer windows, average disagreements exceeded 5%. Even at lower variability levels, the upper tail (mean + 1 SD) of the method disagreement distributions sometimes exceeded 5%. Taken together, these findings highlight the importance of this methodological choice for reducing measurement error and for improving the interpretability of cardiac timing metrics.