Comparing methods for signal analysis of temperature readings from stove use monitors

Understanding the daily use patterns of traditional and nontraditional cooking technologies is essential for researchers and policy makers attempting to reduce indoor air pollution and environmental degradation from inefficient cookstoves. This paper describes field methods and proposes a new algorithm for converting temperature data generated from stove use monitors into usage metrics for both traditional and nontraditional stoves. Central to our technique is recording the visual on/off status of a stove anytime research staff observes the stove. The observations are regressed against temperature readings in a logistic regression to estimate the probability that a temperature reading indicates usage. Using this algorithm we correctly predict 89% of three stone fire observations and 94% of Envirofit observations. The logistic regression correctly classifies more observations than published temperature analysis algorithms. This is the first published algorithm for converting temperature data for traditional stoves such as three stone fires.