Katkat
Back

Estimate weather’s effect on mental health

GoogleData Scientist · HR Screen
EasyData ScientistHR ScreenStatistics & Math

Last updated

This problem tests the ability to design an observational study to measure the causal impact of meteorological variables on self-reported mental health scores, focusing on confounding variables and identification strategies.

Problem Statement

You are working for a health-tech platform that collects daily self-reported 'mood scores' (1–10) from users globally. Your product team wants to understand the causal impact of local daily temperature on user mental health. You have access to the following datasets:

  • user_mood: A daily log of user ID, date, location (zip code), and mood score.
  • weather_data: Daily logs of temperature, precipitation, and sunlight hours per zip code.

Design a statistical framework to estimate the effect of temperature on mood. You must address the following:

  1. Define your treatment variable and the target outcome variable.
  2. Identify at least three potential confounding variables that could bias your estimate.
  3. Propose a statistical model or estimation strategy to isolate the causal effect of temperature.
  4. Describe how you would validate your findings to ensure the relationship is not merely correlational.

Hints

Hint 1

Consider how seasonal trends (like the transition from winter to spring) might act as a confounder for both mood and temperature.

Hint 2

Think about using a fixed-effects model to control for time-invariant characteristics of users or locations.

Hint 3

Discuss how you might check for robustness by performing a placebo test, such as using a weather variable that shouldn't affect mood.

Follow-up questions

  • How would you handle missing data if users only report their mood sporadically rather than daily?
  • If you found a non-linear relationship (e.g., mood is best at 22C and drops at extremes), how would you adjust your model?
  • How would you account for the fact that people in different climates might adapt to temperature differently (e.g., a 30C day in London vs. Dubai)?