State space methods for official statistics and climate modelling

In this thesis, multivariate state space models are developed to combine time series observed with repeated sample surveys with a large number of auxiliary series based on related Google searches. To avoid high-dimensionality problems, a dynamic factor model is developed.
Furthermore, a method to allow for time-varying correlations between trends in a seemingly unrelated time series equation model is proposed. Finally, spatial state space models are developed for modelling nitrogen concentrations in the atmosphere.
Schiavoni, C (1992). Multivariate state space methods for official statistics and climate modelling. Dissertation, Maastricht University, doi:10.26481/dis.20211104cs.
Downloads
- PDF - Dissertation Schiavoni