A Comparison of Different Statistical Methods for Addressing Censored Left Data in Temporal Trends Analysis of...

This study compared four different statistical methods, involving six estimation procedures, for addressing censored left data in measuring temporal trends of eight different pyrethroids measured in sediment from a 10-year data set in a residential California stream (Pleasant Grove Creek). The statistical methods used were: the Kaplan-Meier (km) method; the robust regression on order statistics (ros using normal and log normal distributions rosln); the maximum likelihood estimation (mlen using normal and log normal distributions mleln); and a substitution method (sub) using ½ the detection limit. For five of the eight pyrethroids (bifenthrin, cyfluthrin, cypermethrin, lambda-cyhalothrin, and permethrin), the six statistical methods generally agree, with one exception, that the data set exhibit significant declining trends. In the case of bifenthrin, the slight disagreement among statistical methods only occurred for the mleln estimate that did not show a significant declining trend, whereas the other five methods did. For deltamethrin, esfenvalerate, and fenpropathrin, all six statistical methods were in agreement showing no significant trends. Possible reasons for declining sediment concentrations of pyrethroids in Pleasant Grove Creek are urban label changes effective in 2012-2015 that reduced residential use, variable annual rainfall, and more responsible homeowner use based on outreach/education programs.