• hidden Markov model
• multifrequency data
• quantile regression

This thesis explores the Hidden-Markov-Switching Quantile Regression model, a blend of quantile regression and Hidden Markov Models (HMMs). Unlike classical regression that focuses on the conditional mean, quantile regression examines the impact of covariates on various quantiles of the response variable distribution, which is useful in economics and finance. HMMs, which assume that the regime transition of observed data is a finite-state Markov chain, are used to capture changing dependencies in economic variables.
The model is developed both for homogeneous frequency data, and multifrequency data.
Various statistical tests are implemented to assess the model’s robustness and reliability and their efficacy tests is evaluated using simulated data.
A simulation study is included to examine the behavior of the proposed model, presenting results from the estimation of quantiles of GDP for both the EU area and the US. Finally the theoretical model is extended by incorporating a Quantile Autoregressive structure on the response variable and consistency of the quantile regression estimates is proven under reasonable assumptions.