Comparing the accuracy and speed of four data-checking methods.

Abstract

Double entry locates and corrects more data-entry errors than does visual checking or reading the data out loud with a partner. However, many researchers do not use double entry, because it is substantially slower. Therefore, in this study we examined the speed and accuracy of solo read aloud, which has never before been examined and might be faster than double entry. To compare these four methods, we deliberately introduced errors while entering 20 data sheets and then asked 412 randomly assigned undergraduates to locate and correct these errors. Double entry was significantly and substantially more accurate than the other data-checking methods. However, the double-entry participants still made some errors. Close examination revealed that whenever double-entry participants made errors, they made the two sets of entries match, sometimes by introducing new errors into the dataset. This suggests that double entry can be improved by focusing attention on making entries match the original data sheets (rather than each other), perhaps by using a new person for mismatch correction. Solo read aloud was faster than double entry, but not as accurate. Double entry remains the gold standard in data-checking methods. However, solo read aloud was often substantially more accurate than partner read aloud and was more accurate than visual checking for one type of data. Therefore, when double entry is not possible, we recommend that researchers use solo read aloud or visual checking.