Approximate Entropy for Testing Randomness of Measurement
Approximate Entropy for Testing Randomness of Measurement
Jorge Cogno, INTI
This paper explores the numerical calculation of the information entropy embedded in a set of measurement values, and particularly the application of a metrics called approximate entropy (ApEn) and related statistics ApEn(2,r,N). ApEn proves to be useful in revealing the existence of regularity patterns not shown by other techniques such as autocorrelation or power spectrum, by allowing to obtain a single number indicative of the amount of patternness, from orderly to random, present in sequential measurement data. Capabilities of ApEn and autocorrelation of first- and second- order differences (AFOD-ASOD) to detect subtle patternness in real measurement data are put under test, and compared with analytical expressions for IID Gaussian processes.
20.00