Keywords: Mixed causal and noncausal processes, Student-s’t distribution, Time-reversibility, GCov estimator
 

"Essays on univariate and multivariate noncausal processes"

In econometrics, time series analysis utilizes statistical models to understand the behaviour of economic and financial variables over time. It involves studying sequential observations of one or more variables at regular intervals (e.g., daily, monthly) to identify patterns, trends, and relationships. This analysis helps economists and policymakers understand temporal dependencies, fluctuations, and shocks affecting economic variables.

Various models are employed in time series analysis, such as autoregressive processes (AR), moving average processes (MA), autoregressive moving average processes (ARMA), autoregressive integrated moving average processes (ARIMA), and seasonal models like seasonal ARIMA (SARIMA).  However, these models do not take into account how the variables under investigation might be influenced by future events. In this context, there has been significant recent attention given to mixed causal and noncausal processes. Unlike traditional models, these integrated models consider both past information and future data, enabling them to capture unique patterns, including speculative bubbles.

This thesis aims to delve deeper into mixed causal and noncausal models, contributing to the advancement of identification, estimation, and inference issues.

Click here for the live stream.