A study on the theory of parametric identifiability and its applications in new distributions and classes of probability distributions
Measure. Probability. Distribution. Mixture. Identifiability.
Lately, many authors have proposed new classes of distributions, which are modifications of distribution functions that provide hazard functions taking various forms. Several families proposed in the literature constitute generalizations of probability distributions because, in general, the resulting distribution and the baseline have the same support. It is well known that adding parameters to distribution classes can lead to problems with identifiability and consequently bring complications to the estimation of parameters in the proposed model. This work presented definitions, which are contributions of this thesis to the theory of identifiability. As mentioned earlier, some theorems and propositions based on the aforementioned definitions have been introduced to present a new perspective on the theory. Mixtures of probability distribution functions involving addition and multiplication operations were presented and studied in terms of identifiability. The class of distributions T-G-X [Method for generating distributions and classes of probability distributions: the univariate case, Brito et al. (2019)] was displayed and discussed the identifiability of some subcases of the same. The T-G-X constitutes a multi-baseline extension, hence the bold G, of the well-known T-X class. Each G that makes up the vector of baselines G is a
univariate probability distribution function.