A model that is easy to apply is the so-called Waring-Herdan-Muller model [Muller, Charles, 1979, Du nouveau sur les distribution lexicales: la formule de Waring-Herdan, in "Langue Française et Linguistique Quantitative, Geneva, Slatkine]. According to this model, the number of types "n" that occur with token frequency "r" in the sample can be expressed as a function of "r" and two additional parameters, "a" and "x". (When "a" is fixed at zero and "x" at unity, the Waring-Herdan-Muller model reduces to Zipf's law).
- Baayen & Lieber (2004), a pag.96