Capturing the rhythm of time: introducing Dynamic Time Warping in behavioral research

The growing use of high-intensity time series data has increased reliance on lag models to examine delayed effects. However, one key limitation of traditional fixed-lag models is their inability to flexibly account for variation in lagged effects within and across individuals. To address this, the current study introduces Dynamic Time Warping (DTW) as a technique that accommodates between- and within-person variation in the degree of temporal warping , defined as the extent to which time series need to be stretched or compressed in time to optimally align with each other. DTW is an algorithm that aligns two time series by flexibly stretching or compressing them to minimize their overall distance or mismatch. A key advantage of DTW is that it is model-agnostic, making it particularly suited for dealing with variation in temporal dynamics across diverse psychological processes—regardless of the underlying statistical model. To showcase its value, we compare DTW to DSEM in a simulated dataset, and to traditional lagged regression models in two Experience Sampling Method (ESM) datasets.