A quantitative model of cookstove variability and field performance: Implications for sample size. (January 2015)

• Record Type: Journal Article
• Authors: L'Orange, Christian; Leith, David; Volckens, John; DeFoort, Morgan
• Is Part Of: Biomass and bioenergy. Volume 72(2015:Jan.)
• Page Start: 233
• Page End: 241
• Publication Date: 2015-01
• Subjects: Biomass -- Uncertainty -- Numerical model -- Sample size -- Field testing -- Wood combustion; Biomass energy -- Periodicals; Biomass -- Periodicals; Energy-Generating Resources -- Periodicals; Bioénergie -- Périodiques; 333.9539
• DOI: 10.1016/j.biombioe.2014.10.031
• Languages: English
• ISSNs: 0961-9534

Abstract

Abstract: Many cookstove studies conducted in the field fail to measure meaningful differences between different stove technologies. Although meaningful differences do not always exist, significant differences are often missed because of low statistical power. A numerical model has been developed to determine the minimum sample size necessary to ensure that cookstove field studies are well-designed, efficient, and have adequate statistical power to characterize the concentrations of pollutants inside homes. The numerical model uses a Monte Carlo prediction method to generate probabilistic distributions of indoor pollutant concentrations. The model is based on a series of user inputs, including emissions rate, home size, air-exchange rate, fuel-moisture content, and measurement error. Application of this model to an example situation showed that, even under optimistic measurement conditions, a substantially high number of test replicates would be required. This approach should allow organizations to select appropriate sample sizes to test cookstoves in the field and to identify factors that contribute to variability among tests. Highlights: Studies often do not use large enough sample sizes. We developed a numerical method of calculating minimum sample sizes. Field studies need to use larger sample sizes to find statistical results. Understanding how factors effect uncertainty allows better designed studies. Quantifying the field testing uncertainty is needed to interpret results.