The collection of longitudinal data in the social and behavioral sciences has been revolutionized by the widespread availability of information technologies such as smart phones, social media, physiological sensors, online games and simulations, virtual and augmented reality, and the internet more generally. We use the term intensive longitudinal data (ILD) inclusively, to encompass data coming from a broad range of data collection methods and research designs that are characterized by a relatively large number (e.g., > 60 time points) of multivariate observations collected over time from multiple respondents. ILD present challenges in the application of conventional “large N, small T” methods for longitudinal data (e.g., growth curves, panel models) as well as “small N, large T” methods such as time series analysis and signal processing. One challenge in particular that has received relatively little attention in the social science literature is that of forecasting constructs of interests such as behaviors (e.g., substance abuse) and emotions (e.g., depressive states). Forecasting allows for real-time inferences to be made on the basis of ongoing data collection, which is a key methodological step toward harnessing the full potential of ILD. The purpose of this special issue is to promote contributions that apply or develop new methods to address the problem of forecasting of future events in ILD.
The types of forecasting applications we have in mind are characterized by the following. This list is intended to be illustrative, not exhaustive.
• Data-based protocols for clinical interventions (e.g., nudges, just-in-time interventions).
• The use of educational technologies to support optimal learning trajectories.
• Predicting emergent dynamical states in social networks, dyads, or individuals.
• Improving real time feedback and adaptivity in human-computer interactions.
• Anticipating heightened physiological reactivity (e.g., stress) before it occurs.
Potential extensions of existing forecasting methodology can be motivated by many characteristics of ILD. For example, the data can be:
• high dimensional and multimodal (e.g., quantitative, textual, audio/visual),
• characterized by both within-person and across-person variability,
• sampled at irregular time intervals,
• sampled using different technologies (self-report, machine-collected), and / or
• subject to various issues involved with data collection from human participants (missing data, measurement error).
Manuscripts published in this special issue will be methodologically rigorous and illustrate the application of innovative forecasting methodology with one or more real data examples of general interest to social, psychological, neural, or behavioral scientists. Manuscripts may provide novel analytic developments (for consideration in the T&M section) or a novel application an existing method (for consideration in the ARCS section). Junior scholars are especially encouraged to submit their projects.