Stable estimate of OC/EC ratios

Abstract

In scientific studies, sample ratios are often used to estimate population totals when the dependent and independent variables are believed or assumed to be proportional, such as the primary organics and elemental carbon particles. However, measured data often have errors which can arise from natural variability, sampling errors, instrument limitations, difference in analysis methods, etc. Statistically, a linear regression of Y on X would be a better approach to estimate the average ratio of Y/X, particularly when the ratio of Y/X is not constant over the range of X. However, in regular regression X is assumed to have no error which is not true in reality. In this paper, numerical tests are carried out to examine influence of errors in both variables on the estimation of ratios. The results are applied to the estimation of secondary organic aerosol contribution to PM2.5 concentrations using the EC tracer method.