Regression models for ordinal valued time series: applications in high frequency finance and medicine

In this paper we investigate intraday data of the IBM stock and a time series representing the sleep states of a newborn child. In both cases we are interested in the influence of several covariates observed together with the response series. For the purpose we use on the one hand the regression model proposed in Müller and Czado (2002), on the other hand the ordered probit model. The parameters are estimated with the GM-MGMC algorithm described in Müller and Czado (2002). Predictions are computed to test the results.