JMIR Publications

ABSTRACT

Digital data contains important information about both the behaviours people perform, and the time-of-day they perform these. For example, data collected from social media platforms can include precise measures of when individuals engage and interact with these platforms. Given the political and public interest around the impact of nighttime use of social media platform, effectively collecting and analysing this data is essential. This requires approaches which can take large-scale and complex social media data, and produce accurate and useful measures of the time-of-day these platforms were used. However there is little guidance available to researchers around which approaches are currently available, and when they should be used. The purpose of this tutorial is to demonstrate and evaluate the variety of approaches that have been used to quantify the time-of-day individuals use social media platforms, and how such research can be carried out using real-word data. Our tutorial aims to be a detailed guide for researchers interested in using social media data to explore questions around the timing of behaviour. In this tutorial, we used data collected from TikTok, which contains around 10,000 TikTok videos, watched by a single user between 2025 and 2026, and the precise time-of-day when each was watched. In our tutorial, we provide detailed instructions on how such data can be collected and prepared for analysis. Next, we use five different approaches to calculate estimates of the time-of-day TikTok videos were watched by this user. These approaches focus splitting hours into pre-determined time windows, using methods from the field circular statistics or calculating the seconds that pass between two consecutive social media behaviours. We provide detailed instructions on how each approach can be implemented, and evaluate their strengths and limitations. Finally, we provide recommendations on which approaches should be used for particular types of social media data. Ultimately, these recommendations depends on the type of data collected, and which research questions are of interest.