Paulino and de Bragan<U+00E7>a Pereiria considered 97
subjects who each were one of three risk types: “low”,
“medium”, and “high” (or 1, 2, 3).Of the 97 subjects, 51 were fully categorized: 14 were type 1, 17 type
2, and 20 type 3; the remaining 46 were partly categorized: 28 were of
types 1 or 2, and 18 were of types 2 or 3.
The likelihood function is thus
$$
p_1^{14}p_2^{17}p_3^{20}\left(p_1+p_2\right)^{28}\left(p_2+p_3\right)^{18}
$$
and object paulino gives the PDF.
Note the object gives the likelihood, not the posterior
PDF. Paulino and de Bragan<U+00E7>a Pereira use
prior information of a type not readily representable in the
hyperdirichlet paradigm.