Precision of arsenic summary measures
An 8-mile radius was required to link all homes to a publicly monitored water source with usable arsenic data. At this distance, the median number of arsenic samples was 105 (range 1-507) from 1-281 (median 48) water sources. At much shorter radii, the number of residences linking to arsenic-sampled water sources was modest, but precision improved greatly (Table 2). Using a half-mile radius, agreement was acceptable and similar across summary measures (ρ = 0.47-0.51), but only 30% of homes were included. Doubling the radius (1 mile) doubled the homes included (58%), but agreement dropped notably (ρ = 0.26-0.36). Agreement at larger radii was poor to modest (ρ = 0.04-0.32). At radii sufficient to include all homes, use of only the most recent monitoring records maximized agreement.