The standard paradigm for statistical modelling assumes that observations on variables are independent. In addition to multivariate statistical techniques that afford for dependencies between variables, and time series analysis with dependence over time, many empirical settings give rise to complex data structures that place additional requirements on modelling and inference.
These types of dependencies include:
- Spatial and temporal dependencies
- Group dependencies
- Peer effects and other network dependencies
Common modelling frameworks for these types of data include
- Markov random fields
- Gaussian processes
- Graphical models
- Marked point processes
- Hierarchical and multilevel models
as well as are-specific approaches such as Exponentail Random graph models and Stochastic Actor-oriented models, for network data.