Noisy radioactivity data analysis using parametric Poisson models
Résumé
In metrology, the use of decision threshold and detection limit concepts often creates many
problems for metrologists in radioactivity analysis laboratories. It is usual to censor data when
it becomes difficult to discern the presence or absence of the activity, due to noise in the measurement data. This implies that if the measurement results are non significant which are below
critical value of the test statistic which is called the decision threshold (DT) in metrology. The
analysis simply states that the true value (signal) of the radioactivity is below a certain limit
called the detection limit (DL). These problems are frequently related to the incorrect understanding of the DT formulas or the wrong choice between several formulas whose numerical
results are significantly different. Moreover, it is often unclear how to generate an appropriate and justified DT. In the current research paper, we elaborate a statistical method of DT
determination, capable of providing DT with a high statistical power, using a smaller number
of repeated measurements. The method is then applied to a real test case. Next, statistical
approaches methods are adopted to estimate the density, the expectation and the variance of
the radioactivity. Some of its asymptotic properties are also discussed. Efficiency and feasibility
of these approaches are corroborated through applications on simulated real data sets.
Origine : Fichiers produits par l'(les) auteur(s)