Modeling the term-structure movements of interest rates is a task that has been attracting a crescent number of researchers and practitioners in quantitative finance, given its importance as the main driver for the economic cost of capital. The volume of traded interest rate sensitive assets and derivatives has grown significantly over the last few years, followed by increasingly complex models of pricing and analysis. The high dimensionality of the object of study requires the use of mathematical tools quite different from standard stock market models, resulting in several approaches that eventually lack a unified framework, flexible enough to capture the dynamics of some particular markets. In Brazil the yield curve analysis is even more complex, due to the fast increase of fixed income products over the last ten years, and the historical shifts in the monetary policy conduction. The risk premium in the Brazilian term-structure of interest rates is partially driven by some specific defensive behavior, following past monetary decisions. Until 2008, the Brazilian Central Bank has primarily dealt with domestic and external crises by raising the short term rate to restrain capital outflows, generating a well-known asymmetry in the market’s response functions to risk aversion. Therefore, the traditional parameterization of risk based on mean and variance estimators fails to capture the market price of risk assigned to higher order moments of bond returns across several maturities. The main purpose of this thesis was to get a broad picture of the singularities of the Brazilian term-structure dynamics, and use it to propose alternative approaches to interest rate derivatives pricing – particularly, embodying the third and fourth (pseudo) moments of bond returns into the modeling cycle. The work is divided in two parts: the first exploratory, applying multivariate statistics, portfolio theory and risk management tools to trace the historical evolution of the Brazilian yield curve, and plot the timeline of risk premia and prices of risk linked to exogenous and endogenous factors. The second part of the research uses the statistical evidence gathered as input to a semi-parametric model for pricing derivatives, based on elements of Information Theory. The model was back-tested over an extensive database of local interest rate options, and compared to the results of a traditional market model (BGM). The thesis concludes that the dynamics of the Brazilian yield curve is in part driven by its historical heritage, and endogenous risk factors including moments of bond returns of third and fourth orders are relevant for the premia structure and evolution. Bringing these elements into a modeling process might partially bridge the gap between classical curve models and the local pricing practice.