Impact of COVID-19 on public social life and mental health: A statistical study of Google Trends data from the USA
Archi Roy, Soudeep Deb, Divya ChakarwartiJournal: Journal of Applied Statistics.
The authors evaluate how the public mental health has been impacted by the recent COVID-19 pandemic in some selected states of the USA. They utilize Google Search Volume (GSV) data for keywords belonging to nine categories which are related to positive and negative thoughts, psychological issues, health, hygiene, hobbies etc. After a detailed empirical investigation and keeping in mind the aim of the study, they analyze data corresponding to each state-keyword combinations separately. For a particular state, they first detect the timepoints at which the searching trends in each category underwent a significant change during the pandemic. Then, the data is analyzed via an appropriate time series modelling technique, where the focus is on estimating the effects of the changepoints, and the severity of the pandemic. They assume a general structure of the time series model to allow efficient estimation of the effects.
Their methodology identifies that the changepoints in the searching trends in different categories and for all states typically align with the three waves. For all keywords, they find that the search trends have been significantly impacted by at least one wave of the pandemic, although the effect sizes are different across states. Another key finding is that the impact of the severity of COVID-19 is usually non-significant, indicating that the mental health has been generally more impacted by the long duration of the pandemic rather than the severity of it.
Their analysis further suggests that the model not only explains the variability of the data in a better way, but also renders good predictive performance. They believe that the findings can help them decipher the changes in people’s behavior during a similar crisis. Moreover, they show that the modelling procedure works fairly well under more generalized setup and hence can be applied to similar data structures as well.
