Application of the β-substitution method for analyzing censored data on indoor radon concentrations

Abstract

When analyzing the results of radon surveys, there could be an issue with measurements that are below the limit of detection (LOD) of the measurement technique used. This issue is usually referred to as the management of left-censored data. The paper presents the results of the practical application of one the most accurate and easy to calculate methods for analyzing censored data – the β-substitution method – to the analysis of an indoor radon concentrations dataset. The results of the radon survey conducted in the framework of the regional program “Radon” in Krasnoselskiy district of St. Petersburg in 2000 have been used as the test dataset. It is shown that inappropriate approaches to the management of the censored data, such as discarding the censored values or substituting them with the LOD value, can lead to considerably biased estimates of parameters or statistics of the resulting distributions. Further, this could result in an overestimation of doses and risk estimates which are based on the median values. The visual and quantitative analysis of Q-Q plots leads to conclusion that the estimated parameters of the distribution after the application of the β-substitution method are characterized by a minimum bias. The applied β-substitution method could be recommended for use as an element in the procedure of analyzing the results of radon surveys to minimize the bias of the estimates of the parameters of the radon concentration distributions alongside with Q-Q plots used to verify the conformity of radon concentrations with a lognormal distribution.