INFORMATION LOSS DUE TO DATA QUANTIZATION IN RECONSTRUCTABILITY ANALYSISf

Publisher Website
Google Scholar
Abstract
When the methods of Reconstruct ability Analysis, in any of its several probabilistic forms, are applied to systems of continuous-valued variables, the data must be quantized before (he computer analysis is begun. The question of how coarsely or finely to quantize the data has so far been an open question, and this paper explores the effect different degrees of quantization have on (1) the loss of information incurred about the original system, (2) the amount of data required for statistically reliable reconstructability analysis, (3) computational requirements of the analysis, and (4| the reliability of indirect significance testing of Transmissions through a chi-squared approximation. Coherent conclusions are reached from several perspectives regarding optimal degree of quantization for rcconstruclability analysis.

This publication has 5 references indexed in Scilit: