Data Validation Study.

Abstract : Research was conducted and methods were developed for detecting and locating errors in large data systems. Errors were assumed to originate from any source including, recording, transcription, man-made, machine-made, etc. Mathematical formalism was developed to evaluate probabilities calculated from the results of tests applied to the data. The tests are based on redundancies and known relationships in the data, as well as past history of the data. The logical implication of sets of failed tests are evaluated for various models including dependency and independency of errors assumptions. Methods were developed for the construction of dictionaries of tests and procedures given to check the dictionaries for consistency, completeness and redundancies. The effectiveness of tests to detect and locate errors was defined for single tests and sets of tests giving a way to evaluate the testing procedure. (Author)