This article discusses some modifications to non arbitrage models described by an Ornstein- Uhlenbeck latent process and an affine dynamic system. The empirical analysis refers to Euribor rates, due to the leading role they have in financial markets, but also to help the replicability of the results due to their accessibility and gratuitous nature. The benchmark model belongs to the class of Affine Term Structure Models (ATSM), whom owe their popularity to the success of Duffie and Kan (1996). Nodes have been calculated recursively through the use of the Kalman filter, and hence have the corresponding bayesian interpretation. The proposals differ from traditional models on some constraints posed on certain model specifications that allow to identify different aspects of the term structure. Through a clear identification of the type of contribution that each factor can undertake, it is possible to define probabilistic structures with minimal residuals purified from the dominant systematic residues visible in classic model residuals. Term Structure properties seem to be be identified with greater precision, which in the authors opinion justifies the relaxation of the hypothesis due to the additional constraints. The empirical analysis tries to convey such findings, and reminds of possible evolving paths of this line of work, such as a different specification of the transition process or the relaxation of linear and gaussian nature.
