This thesis consists of four essays on nonlinear approaches in macroeconomic modeling. The first essay analyzes a structural vector autoregressive model to assess the contribution of inflation targeting monetary policy to economic stability. This study is conducted with the data of New Zealand, which was the first country to officially adopt inflation targeting. It is shown that the inflation targeting policy contributed to the overall macroeconomics stabil-ity. In the second essay, the expectation-maximization (EM) method, which is adapted to regime-switching settings, is derived. In the third essay, the perfor-mance of the EM method is tested by replicating well-known examples from the literature on regime-switching models. The method is shown to provide a good trade-off between accuracy and speed for a setting involving large data sets available in mixed frequencies. In the fourth essay, the performance of the EM method is tested using real-time US data. Here, a linear nowcast-ing model is used as a benchmark and compared to a modified framework involving regime-switching. The nowcasting performance of both models, the benchmark and regime-switching, between 2007 and 2022 is evaluated. The regime-switching model is shown to outperform the linear model in terms of short-term prediction performance, especially during the COVID-19 pandemic.