Latent Markov Factor Analysis: A Mixture Modeling Approach for Evaluating Within- and Between-Person Measurement Model Differences in Intensive Longitudinal Data

SUMMARY DOCTORAL DISSERTATION: Experience sampling methodology, in which participants are repeatedly questioned via smartphone apps, is popular for studying psychological constructs or “factors” (e.g., well-being or depression) within persons over time. The validity of such studies (e.g., concerning treatment decisions) may be hampered by distortions of the measurement of the relevant constructs due to response styles or item interpretations that change over time and differ across persons. In this PhD project, we developed a new approach to evaluate person- and time-point-specific distortions of the construct measurements, taking into account the specific characteristics of (time-intensive) longitudinal data inherent to experience sampling studies. Our new approach, latent Markov factor analysis, extends mixture factor analysis and clusters time-points within persons according to their factor model. The factor model describes how well items measure the constructs. With the new approach, researchers can examine how many and which factor models underlie the data, for which persons and time-points they apply, and thus which observations are validly comparable. Such insights can also be interesting in their own right. In personalized healthcare, for example, detecting changes in response styles is critical for accurate decisions about treatment allocation over time, as response styles may be related to the occurrence of depressive episodes.

Dynamic models of stress-smoking responses based on high-frequency sensor data

Self-reports indicate that stress increases the risk for smoking; however, intensive data from sensors can provide a more nuanced understanding of stress in the moments leading up to and following smoking events. Identifying personalized dynamical models of stress-smoking responses can improve characterizations of smoking responses following stress, but techniques used to identify these models require intensive longitudinal data. This study leveraged advances in wearable sensing technology and digital markers of stress and smoking to identify person-specific models of stress and smoking system dynamics by considering stress immediately before, during, and after smoking events. Adult smokers (n = 45) wore the AutoSense chestband (respiration-inductive plethysmograph, electrocardiogram, accelerometer) with MotionSense (accelerometers, gyroscopes) on each wrist for three days prior to a quit attempt. The odds of minute-level smoking events were regressed on minute-level stress probabilities to identify person-specific dynamic models of smoking responses to stress. Simulated pulse responses to a continuous stress episode revealed a consistent pattern of increased odds of smoking either shortly after the beginning of the simulated stress episode or with a delay, for all participants. This pattern is followed by a dramatic reduction in the probability of smoking thereafter, for about half of the participants (49%). Sensor-detected stress probabilities indicate a vulnerability for smoking that may be used as a tailoring variable for just-in-time interventions to support quit attempts.

Modelling agreement for binary intensive longitudinal data

Devices that measure our physical, medical and mental condition have entered our daily life recently. Such devices measure our status in a continuous manner and can be useful in predicting future medical events or can guide us towards a healthier life. It is therefore important to establish that such devices record our behaviour in a reliable manner and measure what we believe they measure. In this article, we propose to measure the reliability and validity of a newly developed measuring device in time using a longitudinal model for sequential kappa statistics. We propose a Bayesian estimation procedure. The method is illustrated by a validation study of a new accelerometer in cardiopulmonary rehabilitation patients.

Perceived vaccine safety over time in a vaccine hesitant sample: Impact of pausing due to safety concerns

In the current effort to vaccinate as many people as possible against COVID-19, it has been suggested that events such as the pause in the use of the Janssen vaccine would have a large effect on perceptions of vaccine safety. Further, as vaccination rates slow, there is concern that hesitancy may be stable and difficult to change among those still unvaccinated. We examined both of these issues in our ongoing study of low-income participants. We modeled the intensive longitudinal data provided by 53 individuals. We found the negative, not statistically significant effect of the Janssen pause would be overwhelmed within weeks by the statistically significant increasing perceptions of safety across time. We also observed strong variability in vaccine hesitancy in many participants. Frequent reminders about vaccine availability might catch more people when they are less hesitant, helping increase vaccination rates.

Dynamical Outcome Modeling of Fragmented Intensive Longitudinal Data

The advent of new sensing technologies has enabled the collection of high frequency individual data. Not only can this allow for the identification of personalized behavior models, it also opens up the possibility of developing just-in-time interventions that leverage the information collected to determine when and which micro intervention should be provided. However, there are significant challenges in the analysis of this type of data. First, due to the high rate of data collection, one can no longer assume that the stimulus (independent excitation or micro intervention) only has an instantaneous effect on the outcome, one has to also allow for delayed effects. Moreover, one is also frequently faced with fragmented data; i.e., poor placement of sensors, non-wear of the data collecting device and/or external disturbances can lead to intervals of time where the data collected is not reliable; i.e., missing or corrupted. To deal with these challenges, we leverage concepts from the areas of dynamical systems and signal processing to develop tools that i) can identify models that take into account the delayed stimuli effects and ii) are able to handle fragmented data. In this paper, we provide both the mathematical foundation of the tools proposed and the description of a package that implements them. We also discuss ways to interpret the results obtained.

Personalized Behavior Prediction: An Idiographic Person-Situation Test

A longstanding goal of psychology is to predict the things people do, but tools to accurately predict future behaviors remain elusive. In the present study, we used intensive longitudinal data (N = 104; total assessments = 5,971) and three machine learning approaches to investigate the degree to which two behaviors – loneliness and procrastination – could be predicted from past psychological (i.e. personality and affective states), situational (i.e. objective situations and psychological situation cues), and time (i.e. trends, diurnal cycles, time of day, and day of the week) phenomena from an idiographic, person-specific perspective. Rather than pitting persons against situations, such an approach allows psychological phenomena, situations, and time to jointly predict future behavior. We find (1) a striking degree of prediction accuracy across participants, (2) that a majority of participants’ future behaviors are predicted by both person and situation features, and (3) that the most important features vary greatly across people.