Graduate Student Seminar

Speaker 1: Mark Ramos

Title

Estimation of Multiple Time Series with Volatility

Abstract

The presence of volatility in multiple time series presents a significant modeling problem. In this study, a block bootstrap procedure for modeling single time series with volatility was combined with a backfitting algorithm for modeling multiple time series with common autoregressive patterns in order to address the presence of volatility in the multiple time series case. Simulation studies conducted showed that the hybrid algorithm provided robust estimates of model parameters and accurate predictions of the multiple time series values during non-volatile periods.  The hybrid algorithm was found to be superior to the original backfitting algorithm in terms of MAPE and parameter estimate bias in most settings when volatility was present, except when the multiple time series length was very short or when the autocorrelation parameter used was close to 1.

Speaker 2: Guy Djokam

Title

A model for TB with exogenous infection and co-infection with HIV

Abstract

The problem under interest in this dissertation concern the modeling and mathematical and numerical analysis of an epidemiological model applied to tuberculosis. A particularity of this model is that it taking into account exogenous infections and co-infections with HIV on the propagation of the disease. We present a result on the computation of threshold parameter Ro  and it relation with the global asymptotic stability of the disease free equilibrium. The effectiveness of the obtained analytical results has been illustrated by numerical simulation.