Regression models for binary time series

Specialeforsvar ved Franziska Juchmes

Titel: Regression models for binary time series

Abstract: In this thesis, we want to investigate a time series taking only binary value. We will differentiate between an observation and parameter-driven model. Our focus will be on parameter-driven model. The model is going to include a latent process and use a logistic model to reflect the covariate-responds relationship in the latent process. As for parametric-driven model the estimation of the likelihood function is not easy. Based on this, we want to estimate the regression parameters from the marginal distribution also derive the parameters for the latent process with the help of some composite likelihood methods. We use the bootstrapping method to calculate confidence intervals. 


Vejleder:  Helle Sørensen
Censor:    Sören Möller, SDU