Twitter as a means to study temporal behaviour.

Biomedical research has exploited vital and other statistics (e.g., birth or death rates) for almost 200 years [1]. The Internet has become a rich source of digital databases, which are being used for many lines of research (e.g., circadian and seasonal [2] or metabolism [3,4]). Internet-based studies generally investigate large populations while individual social media accounts are rarely used to analyse, for example, individual sleep-wake behaviour (e.g., youtu.be/wBNcP-LkpfA). I therefore applied time series analyses, commonly used in circadian and sleep research, to approximately 12,000 tweets sent from a single Twitter account (@realdonaldtrump; December, 2014 to March, 2017). The account was clearly used by different individuals/groups launching tweets from various devices. Among these, the Android phone was the most consistent over the years. Its tweet activity peaked twice a day (early morning and late night), and both peaks showed a strong seasonality by tracking dawn.