Abstract |
The paper discusses statistical aspects of autoregressive models in the assessment of radon mitigation. Radon values, as reflected by hourly scintillation counts, seem dominated by major, pseudo-periodic, random fluctuations. The methodological paper reports a moderate degree of success in modeling these data using relatively simple autoregressive-moving average models in order to assess the effectiveness of radon mitigation techniques in existing housing. While accounting for the natural correlation of successive observations, familiar summary statistics such as steady state estimates, standard errors, confidence limits, and tests of hypothesis are produced. The Box-Jenkins approach is used throughout. In particular, intervention analysis provides an objective means of assessing the effectiveness of an active mitigation measure, such as fan off/on cycle. Occasionally, failure to declare a significant intervention has suggested a diagnosis of the data collection procedure. |