Why and How to Deal With Diurnal Cyclic Patterns in Ambulatory Assessment of Emotions

Abstract

Abstract. The use of ambulatory assessment (AA) based methods in emotion research has steadily increased over the past decades. Although having a number of benefits over other methods, the use and analysis of AA data may pose specific challenges. Among these, the issue of dealing with diurnal cycles in emotion data has received relative scant attention. This article therefore discusses why cyclic models may be considered for analyzing AA data on emotions, and describes how this approach can be applied to an empirical AA dataset. Results suggest that cyclic modeling may be a useful method for describing and accounting for (diurnal) cyclic patterns in AA data, but should be used with a number of considerations in mind.

Daily dynamics and weekly rhythms: A tutorial on seasonal ARMA models combined with day-of-week effects

“…We may distinguish three lines of research in psychological literature on analyzing intensive longitudinal data using diary data: (a) some researchers (e.g., Larsen & Kasimatis, 1990;Ram et al, 2005;van de Maat et al, 2020;West & Hepworth, 1991, among others) have modeled the DOWEs in the mean structure (by means of dummy variables for each day or the weekend, or a harmonic wave) alone without considering lagged dynamics; (b) a large body of affective dynamics research in the past decade has focused on lag-1 dynamics, particularly with various autoregressive models, and ignored possible regularities due to the DOWEs in the mean structure and other forms of dynamics; and (c) very few papers (to our knowledge, only Keller & Meier, 2023;Liu & West, 2016; see also Muthén et al, 2024, for within-day cycles) have considered modeling the DOWEs and AR(1) dynamics together. As we demonstrated using a rich daily diary dataset, these three approaches may not apply to a vast majority of individuals: 82% of the individuals within our sample showed some sort of lagged dynamics (which the first approach overlooks); 83% of the individuals showed DOWEs and/or non-AR(1) dynamics (which the second approach cannot accommodate); and 62% of the individuals had moving average and/or weekly dynamics (which fall beyond the capabilities of the third approach).…”

“…An obvious approach for this is based on using a separate dummy variable for each day of the week, which allows for any pattern over the week. Alternatively, researchers have used a sine wave to model a more smooth ebb and flow in daily affect ratings (Beal & Ghandour, 2011;Ram et al, 2005;van de Maat et al, 2020). Such modeling approaches can then also be combined with techniques that account for day-to-day dynamics; for instance, Keller and Meier (2023) and Liu and West (2016) combined dummy variables with a first-order autoregressive process where today's score is partly predicted from yesterday's score.…”

Accurate Prediction of Momentary Cognition From Intensive Longitudinal Data

“…Conditional models additionally included linear (see S1), quadratic (age), cyclic (sine, cosine of time) 72 , and two-way interaction (age, linear x time, cyclic) fixed effects. Change (∆) in variance explained (R 2 ) between participant mean and conditional models reflected the extent to which fixed effects improved estimates of momentary cognition, principally by explaining variation in intra-individual cognitive fluctuations.…”

Negative Emotion Differentiation Predicts Psychotherapy Outcome: Preliminary Findings

“…Specifically, a time scale (e.g., Adolf et al, 2021 ) appropriate for capturing affective changes as they unfold in patients’ daily life should be identified based on prior research (e.g., Verduyn et al, 2009 ) or theoretical grounds, and dictate the measurement scheme. Additionally, trends (e.g., linear and quadratic; Jebb et al, 2015 ) and cycles (e.g., diurnal and weekly; van de Maat et al, 2020 ) should be modeled and interpreted on a case-by-case basis ( Fisher and Newman, 2016 ).…”