New Details

Time series are data points collected at regularly spaced time intervals. Time series of counts are a special type of time series considering only non-negative integer values. An example is the daily number of hospital admissions, which can be zero or any other positive integer value. This type of data requires models that ensure the integer nature of the response variable (hospital admissions), which is not possible with continuous time series models.

Scientific advances in integer-valued models have been prolific in the last decades but focus mainly on univariate models, i.e., models with just one response time series. Thus, we propose a new class of models, called Space-Time Integer-Valued Autoregressive and Moving Average (STINARMA) models. This new approach allows to model, for instance, hospital admissions at several locations simultaneously while accounting for their spatial dependency. These models keep the integer nature of the data by using an operation called binomial thinning and integer-valued innovations. Due to its space-time nature, the STINARMA models can describe complex real-life phenomena in a more accurate way.

Authored by: Ana Martins

In collaboration with: Sónia Gouveia, Manuel Scotto, Christian Weiss

Funded by: Fundação para a Ciência e a Tecnologia (FCT): PhD Grant SFRH/BD/143973/2019, IEETA/UA (UIDB/00127/2020), CEMAT/IST/UL (UIDB/04621/2020)

Paper Reference