Researchers administered a detailed survey to 421 households. In a sub-sample of 36 households, researchers collected 24-h integrated PM2.5 samples (personal and kitchen area); in addition, the primary cook was monitored for real-time PM2.5. All filters were also analyzed for black carbon using a multi-wavelength reflectance method. Predictors of PM2.5 exposure were analyzed, including cooking behaviors, fuel, stove and kitchen type, weather, demographic factors and other smoke sources