Sur une nouvelle application de la théorie de 1'information à la description des séries chronologiques planctoniques

Publisher Website
Google Scholar
Abstract
By the estimation of the probability that a “turning point” occurs at random in a chronological series, we can easily calculate a quantity of information related to each observation, without knowing the exact numeric values of the variable. We can construct a new set of data containing the information quantities, which is adequate to test the validity of the sampling: formal selection of the descriptors useful for the interpretation, detection of series in homogeneous periods. The global evolution of the information defined by the cumulative information of each descriptor, is similar to the classical concepts of diversity, or to the parameters of Motomura's model. Moreover, if the well-sampled stations are favoured by a weighting based on the quantity of information, the efficacy of the treatments can be improved without the loss of the statistical principles. At last, this technique can be more generally applied to a spatio-temporal sampling in two dimensions.

This publication has 0 references indexed in Scilit: