Statistical Analysis of Delay in Time Series

Abstract

This thesis focuses on delay in time series data. The first delay involves the m-delay autoregressive model. This approach considers only the first and the last previous observation of the traditional autoregressive model. Next, the delay is added to the stochastic differential equation for matching the volatility between real-world financial data and Monte Carlo simulations. Finally, a two-delay combination method is proposed to increase the prediction accuracy of the individual deep learning model.