Test if social users are more engaged

A PM has a hypothesis: **users who use “social” apps are more engaged on a regular basis than users who use “game” apps.** You have the same tables: - `user_activity(user_id, date, app_id, session_id, duration)` - `apps(app_id, app_name, app_category)` ## Task Describe how you would **evaluate** this hypothesis using historical data. Your answer should cover: 1) How you would define **engagement** (choose a primary metric and 1–2 diagnostic metrics). 2) How you would define comparison groups when many users use **both** categories (e.g., “social-only”, “game-only”, and mixed users; or thresholding by % time). 3) What statistical test/model you would use to compare groups, and what assumptions you’d check. 4) What you would do if the “social-only” and “game-only” groups are small (e.g., n≈300 each) but the mixed group is very large (e.g., n≈100,000). 5) Key confounders/biases you would worry about and how you’d mitigate them (e.g., user tenure, overall activity level, seasonality).