Statistical analysis
We compared each summary measure to respective contaminant levels in participants' water. Specifically, while retaining these as continuous measures, we estimated precision (hereafter also "agreement"). Because the summary measures we examined likely would be used in lieu of the gold standard (i.e. this was an inter-method comparison), and because contaminant levels were non-normally distributed, we estimated Spearman's correlation coefficient (ρ) as a measure of precision [16]. When ρ = 1.0, the measure to be validated orders all observations perfectly in comparison to the gold standard (i.e. relative exposure levels are fully preserved), and when ρ = 0 there is no relationship between the compared values. We do not report p-values because the magnitude of the correlation is of interest: This estimate is useful in the design and interpretation of studies that include the respective exposure measure. For example, when ρ = 0.50, only very strong associations remain detectible [16]. For our results we defined ρ ≤ 0.40 as unacceptable because with this level of precision even strong associations become very difficult to observe.