An Evaluation of Data Entry Error and Proofing Methods for Fisheries Data

Abstract

AbstractReducing data entry error has the potential to improve estimates produced by fisheries practitioners. However, the frequencies of data entry error and evaluations of the recommended protocols for dealing with data entry error have rarely been presented in fisheries‐related literature. The objectives of our study were to determine the magnitude of data entry error in a typical fisheries data set, what kind of errors occurred most often, and how those errors might affect commonly generated estimates of abundance, size structure, and species richness. We evaluated four methods of data entry and proofing: (1) a single entry, (2) read‐aloud proofing, (3) double‐entry proofing, and (4) field use of a personal digital assistant (PDA). We determined the quality of the data after the use of each method and compared common fisheries estimates derived from each with estimates generated from standardized data. Total error discovered in the data set averaged 0.79 ± 0.22% (mean ± SD) and consisted of 44.1% field‐related errors and 55.9% data entry errors. We found that numbers of known errors remaining in the data were significantly lower when proofing methods were used. Abundance estimates derived from a single data entry were significantly different from those derived from data that had undergone proofing. However, the magnitude of the difference (2.22%) was less than our limit of acceptable error and far less than the mean confidence interval of the estimates themselves (60.91%). Further, no differences were detected in mark‐recapture abundance estimates, estimates of size, or estimates of species richness. This suggests that for most common fisheries estimates, a single entry of data or single entry using a PDA is sufficient. We subsequently found that the use of automated error checking helped to ensure an acceptable level of data quality without the time and expense of more traditional error‐checking methods.