Detecting Data Error and Inaccuracy

Abstract

Several studies reveal that organisational databases have significant errors (Klein, 2000; Morgenstern, 1963; Musgrove, 1974). Attributed to this, researchers, policy makers and other users are obliged to authenticate collected data for its randomness before usage in order to prevent problems caused by erroneous data. The objective of this study is to establish random databases for furthering scientific analysis. Motivation for the study comes from the circumstances where scientific inquiry is juxtaposed with different databases or sources of the same unit of inquiry having different datasets. This study uses three databases used by the Indian government, namely, the Economic Survey, Reserve Bank of India and United Nations Conference for Trade and Development (UNCTAD) on FDI inflows to the Indian economy (1991–2007), to establish randomness and detect data errors and inaccuracies. Applying the Run test method, the study established that all three datasets are random. The Chi-square test supports the dataset used by the Economic Survey and not other databases. Also, the Economic Survey dataset follows Benford’s distribution and its Pearson correlation coefficient is higher than the other sources of data. In general, the Economic Survey database does better in terms of data accuracy compared to other datasets for the period of study.