Comment évaluer un nombre de catégories par échantillonnage

This paper describes a method for estimating the number of categories in an urn of unknown composition. From a priori values for the number and frequencies of categories, it is possible to calculate a sampling error and to estimate the number of individuals to draw which would be necessary for decreasing this error beyond a given risk. Successive assays lead to a convergence towards a convenient representation of the urn, in agreement with the fixed risk.