About half of all surveys were completed alone, with a relatively equal balance among the remaining social contexts.
The strongest correlations we found among all the affect scores measured were quite strong and shared by 3 dimensions of affect:
To get a sense of clusters and sub-populations, we created a tool for viewing population-level stats, aggregated per user:
To get a sense of macro effects due to external events (e.g. seasons), we created a tool to show total participation over time:
To get a sense of macro effects due relative to the life of the study, we created a tool to show total participation relative to the start of study:
NOTE: Given such a small dataset with such a tight range of dates, the participation trend per calendar day is not distinct from the participation trend per day in study. In practice, we expect to detect different effects in these two plots.
We created another view of participation trends, with the current user plotted against all others in the dataset:
NOTE: I need to think more about what this plot of (inter|intra)-day response gaps might be telling us. It does show some structure, including three possible clusters. Mainly, I would say that these two dimension capture distinct information.
Credit: Apoth Development, Inc.