Title: Dynamic models with too many zeroes
Authors: Andrew Harvey - University of Cambridge (United Kingdom)
Ryoko Ito - Oxford University (United Kingdom)
Ryoko Ito - Oxford University (United Kingdom) [presenting]
Abstract: Situations are considered in which a significant proportion of observations are identically zero or have been set to zero because they fall below a certain threshold. When the distribution for non-zero observations is continuous, the net result is a point-mass mixture distribution. We propose an integrated approach for dynamic models with censored observations that is straightforward to apply and logically consistent. The model is fitted to daily precipitation data of selected Australian cities.