Uncertainty bounds for metal levels in waste wood
(English)Manuscript (preprint) (Other academic)
Estimates of the mean levels of various metals in wastes are normally presented without any measures of uncertainty, and, if such measures are given, they are almost always based on subjective judgments. Here, we show that adequate statistical techniques exist for calculating uncertainty bounds. If data are sparse, Bayesian statistical methods can be employed to combine expert knowledge with the information provided by actually measuring metal concentrations in wastes. If larger datasets are available, a viable alternative is to use bias-corrected bootstrap intervals to compute uncertainty bounds. To facilitate selection of a method, we illustrate how sensitive the Bayesian inference is to the prior knowledge that is incorporated into the analysis, and how outliers influence the obtained uncertainty bounds. Inasmuch as levels of metals in solid wastes can vary substantially and the underlying probability distributions are often highly skewed, we examine the performance of lognormal models.
uncertainty, Bayesian inference, bootstrap, lognormal, metals, waste wood
IdentifiersURN: urn:nbn:se:liu:diva-86477OAI: oai:DiVA.org:liu-86477DiVA: diva2:577850