Changelog

Changes in data structures

These are heavy changes, which may break existing code. Such changes will not occur in future releases and are due to the early stage of the package.

• Restructured the spflow_model-class (to better handle NA values, missing neighbors, and to enable predictions)
• Renamed and restructured classes
  • sp_multinetwork-class -> spflow_network_multi-class
  • sp_network_nodes-class -> spflow_network-class
  • sp_network_pairs-class -> spflow_network_pair-class

Other changes in arguments etc.

• the spflow_formula uses P_() instead of G_() to refer to variables on the level of OD-pairs
• argument names in spflow()

New functions

• Added pair_cor() method for spflow_model-class and spflow_network_multi-class. The method creates a correlation matrix for origin, destination and OD-pair variables and their spatial lags.
• Added cor_image() to visualize such correlations.
• Added plot() method for some quick diagnostics of the spflow_model
• Added map_flows() to visualize flow data
• Added spflow_moran_plots() for Moran scatter plots of flow-data and residuals of interaction models
• Added compare_results() to quickly compare multiple model
• Added spflow_refit() to efficiently estimate nested models (this functionality is still experimental)

Additional options, features and fixes

• Added support for coordinates in the spflow_network-class
• Added test for parameter space validity
• Added support for in-sample predictions, with methods “TS”, “TC”
• Added support for out-of-sample predictions, with method “TC”
• Added support for non-cartesian and rectangular cases (for details see current working paper at https://www.tse-fr.eu/sites/default/files/TSE/documents/doc/wp/2023/wp_tse_1409.pdf)
• Added handling of NA values during the estimation