Missingness Mechanisms and Patterns in Experience Sampling Method Studies: When Autoregressive Model Estimators Fall Short

Autoregressive (AR) models are popular for quantifying affective inertia (i.e., the resistance of affective processes to change) using Experience Sampling Method (ESM) data. Although missing observations are common in ESM, their impact on the estimation accuracy of the parameters of the AR models remains understudied, especially for more complex missingness mechanisms such as tail-based missingness, where more extreme affect scores are systematically missing, and missingness across multiple measurements. We investigate the effect of compliance level, specific missingness mechanisms (i.e., missing completely at random versus tail-based missingness), and temporal patterns (whether or not missings occur consecutively) on the bias and precision of the estimate of the autoregressive effect. We show that the presence of missing data increases bias and decreases precision, using either classical or Bayesian estimators. The bias is more severe when compliance is low and, importantly, when the missingness is tail-based. These results urge researchers to make design choices that promote compliance and/or increase the probability of data being missing completely at random: decreasing participant burden by using shorter questionnaires, providing adequate incentives to participants, and considering using planned missingness designs or episode-based sampling.