RESUMO
A fast-growing body of evidence from experience sampling studies suggests that affect dynamics are associated with well-being and health. But heterogeneity in experience sampling approaches impedes reproducibility and scientific progress. Leveraging a large dataset of 7016 individuals, each providing over 50 affect reports, we introduce an empirically derived framework to help researchers design well-powered and efficient experience sampling studies. Our research reveals three general principles. First, a sample of 200 participants and 20 observations per person yields sufficient power to detect medium-sized associations for most affect dynamic measures. Second, for trait- and time-independent variability measures of affect (e.g., SD), distant sampling study designs (i.e., a few daily measurements spread out over several weeks) lead to more accurate estimates than close sampling study designs (i.e., many daily measurements concentrated over a few days), although differences in accuracy across sampling methods were inconsistent and of little practical significance for temporally dependent affect dynamic measures (i.e., RMSSD, autocorrelation coefficient, TKEO, and PAC). Third, across all affect dynamics measures, sampling exclusively on specific days or time windows leads to little to no improvement over sampling at random times. Because the ideal sampling approach varies for each affect dynamics measure, we provide a companion R package, an online calculator ( https://sergiopirla.shinyapps.io/powerADapp ), and a series of benchmark effect sizes to help researchers address three fundamental hows of experience sampling: How many participants to recruit? How often to solicit them? And for how long?
