CMStatistics 2017: Start Registration
View Submission - CMStatistics
B1019
Title: Network time series Authors:  Kathryn Leeming - University of Bristol (United Kingdom) [presenting]
Marina Knight - University of York (United Kingdom)
Guy Nason - Imperial College, London (United Kingdom)
Matthew Nunes - Lancaster University (United Kingdom)
Abstract: A network time series describes observations collected at nodes on a network over time. This network may be known, or require construction from the time series or other information, such as location. The NARIMA (Network ARIMA) framework will be introduced which, although similar to VARMA modelling, allows for changes in the network structure over time. This framework allows modelling and forecasting of large data-sets using few model parameters. As in univariate time series modelling, it is important to take account of any changes in the trend of the network time series. Methods for removing trend will be discussed and demonstrated using an epidemiological example.