THE USE OF REGRESSION EQUATIONS FOR QUALITY-CONTROL IN A PESTICIDE PHYSICAL PROPERTY DATABASE

Quality control is a crucial aspect of database management, particular ly for physicochemical parameters that are widely used in modeling env ironmental fate processes. Complete rechecking of original studies to verify environmental fate parameters is time consuming and difficult. This paper evaluates an alternative, more efficient approach to identi fying database errors. The approach focuses verification efforts on a targeted subset of entries by making use of the relationship between w ater solubility (S) and soil organic carbon partition coefficient (K-o c). Two regression equations, one selected from the literature and one calculated from entries in the database, were used to evaluate the re asonableness of (S, K-oc) pairs among control compared to the targeted outlier group from a total of 59 pesticides. Our hypothesis was that (S, K-oc) pairs that lay far from the regression line were more likely to be in error than those that fit the regression. Database values we re checked against original studies. Identified errors in the database included coding mistakes, miscalculations, and incorrect chemical ide ntification codes. The error rate in outlier, (S, K-oc) pairs was abou t twice that of pairs that conformed to the regression equation; howev er, the error rate differential was probably not large enough to justi fy the use oi this quality control method. Through our close scrutiny of database entries we were able to identify administrative practices that led to mistakes in the data base. Resolution of these problems wi ll significantly decrease the number of future mistakes.