This paper sets out an approach for modelling univariate time series, including those in which observations are available on a daily basis. An underlying continuous time model is formulated and it is shown that this model has important implications for the way in which a discrete model is set up. It is also shown that the continuous time model allows observations subject to temporal aggregation and irregularly spaced observations to be handled relatively easily. The extension to cases where explanatory variables are to be included in the model is also discussed.