Abstract
A set of data may be ‘coarsened’ as a result of enumerators’ or compilers’ efforts to estimate (or falsify) observations. This type of coarsening typically results in excesses of ‘convenient’ numbers in the data sets, such as multiples of 5 or 10 in decimal number systems, apparent as patterns of periodic unit-width spikes in the frequency distributions.
We report on the development of novel Radial Basis Function neural-network techniques for detecting numerical data coarsened by rounding/estimation (or falsification) and quantifying that rounding/estimation. The objective is to provide an alternative to conventional statistical approaches based on missing-data techniques: data coarsened thus are not actually missing, solely shifted in size. The results show that the neural networks can successfully detect and classify the coarsening in data-sets and, hence, yield insights into the ways in which people count when performing enumeration or other numerical data-compilation exercises.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
References
Chen S, Cowan CFN, Grant PM (1991) Orthogonal Least Squares Learning Algorithm for Radial Basis Function Networks, IEEE Transactions on Neural Networks, 2, 2, March 1991, pp 302 - 309.
Crockett RGM, Crockett AC (1999) Historical Sources: How People Counted. A Method for Estimating the Rounding of Numbers, History and Computing, 9, 1/2/3, 1999, pp 43 - 57.
Crockett RGM, Crockett AC (2000) People and counting: how people count in enumeration exercises, 3 rd European Social Science and History Conference, Vrije Universiteit, Amsterdam, April 2000.
Heitjan DF, Rubin DB (1990) Inference from Coarse Data via Multiple Imputation with Application to Age Heaping, Journal of the American Statistical Association, 85, 410, 1990, pp 304 - 314.
Heitjan DF, Rubin DB (1991) Ignorability and Coarse Data, Annals of Statistics, 19, 4, 1991, pp 105 - 117.
Polhill JG, Weir MK (2001) An Approach to Guaranteeing Generalisation On Neural Networks, Pergamon, Neural Networks 14 (2001), pp 1035 - 1048.
Triastuti E, Crockett RGM, Picton PD, Crockett AC (2002) Neural Network Analysis of Estimated Data, Proc. Eunite 2002, Albufeira, Portugal. 19-21 September 2002; pp 161 - 166.
Turner SJ, Crockett RGM, Picton PD, Triastuti E (2001) Genetic Algorithms for Simulating Counting Behaviour, 19th Biennial Conference on Numerical Analysis, University of Dundee, 26-29 June 2001.
Turner SJ, Triastuti E, Crockett RGM, Picton PD, Crockett AC (2002) Intelligent Techniques for Detecting Estimated and Falsified Data, Proc Sixth Multi-Conference on Systemics, Cybernetics and Informatics. Orlando, Florida, USA. 14-18 July 2002, pp 445 - 450.
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2004 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Triastuti, E., Crockett, R., Picton, P., Crockett, A. (2004). Neural Network Analysis of Estimation of Data. In: Lotfi, A., Garibaldi, J.M. (eds) Applications and Science in Soft Computing. Advances in Soft Computing, vol 24. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-540-45240-9_5
Download citation
DOI: https://doi.org/10.1007/978-3-540-45240-9_5
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-40856-7
Online ISBN: 978-3-540-45240-9
eBook Packages: Springer Book Archive