Handbook of Radon: Properties, Applications and Health

0363425 - ÚI 2013 RIV US eng M - Monography Chapter
Brabec, Marek - Jílek, K.
Dynamic Assessment of Radon Source in Buildings, Based on Tracer Gas Experiment Statistical Modeling.
Handbook of Radon: Properties, Applications and Health. Hauppauge: Nova Science Publishers, 2012 - (Li, Z.; Feng, C.), s. 211-242. ISBN 978-1-62100-177-5
Institutional research plan: CEZ:AV0Z10300504
Keywords : radon * dynamic modeling * functional data analysis
Subject RIV: BB - Applied Statistics, Operational Research
https://www.novapublishers.com/catalog/product_info.php?products_id=23545

In this paper, we describe an improved dynamical model for assessment of radon source in a building taken as a single compartment. The data upon which the model operates consist of typical outcomes of a tracer experiment – that is of simultaneously measured concentrations of radon and tracer gas concentrations (they do not have to have the same time resolution, however). The model has time-varying coefficients and it is casted as a state-space model. Hence, it allows for essentially non- or semi-parametric estimation of time-varying radon entry rate (RER) and air exchange rate (ACH) based on (extended) Kalman filtration. The state space model is carefully formulated so that it produces more or less automatically not only point estimates of both RER and ACH, but also their confidence intervals and/or test of various practically interesting hypotheses, including those related to the radon remedial action limits violations. Compared to our previously published work, the model explicitly accounts for several external influences and effectively corrects for their presence. These are tight to direct weather-related influences influencing pressure balance, as well as to (functional) relation between ACH and RER. To this account, we will present alternative to a “free” model which assumes that true ACH and RER dynamics are independent. It will introduce various types of explicitly specified ACH to RER dependencies, including those being based on linear functionals of ACH. We will use maximum likelihood estimation (MLE) of structural parameters of the state-space.
Permanent Link: http://hdl.handle.net/11104/0199324