Mathematical models of opinion formation have been studied by physicists mainly since the 80s and are now part of the new branch known as Sociophysics. This recent area of research borrows tools and concepts from statistical physics. Models of this kind are providing good results to describe certain aspects of the social and political behavior, such as the formation of opinions, adoption of new technologies or extreme attitudes, that present a phenomenology, e.g., order-disorder transitions, analogous to some physical systems. Within this scenario, fits this thesis. We study a model of opinions that can be associated to any public debate with three options (yes, no, undecided). We consider a fully connected population of individuals (or agents), which can be in three different states. Interactions occur by pairs and are competitive, being negative with probability p or positive with probability 1-p. This bimodal distribution of interactions produces a behavior similar to that resulting from the introduction of contrarians (in the sense of Galam) in the population. Furthermore, we consider that a certain fraction d of individuals are intransigent or stubborn, usally called inflexibles in opinion dynamics. These individuals keep their opinions unchanged. By means of computer simulations, we study the impact of competition between contrarians and inflexibles on the formation of the majority opinion. Our results show that the presence of inflexibles affects the critical behavior of the population only if such condition is quenched, that is, if the intransigents not change their beliefs with time. On the other hand, in the annealed version of the model, where the inflexibles are chosen at each time interval (that is, stubborness is occasional), the nonequilibrium phase transition which occurs in the absence of inflexibles is not affected. We also discuss the relevance of the model to real social systems.