panelPomp-package {panelPomp}

Inference for PanelPOMPs (Panel Partially Observed Markov Processes)

Description

The panelPomp package provides facilities for inference on panel data using panel partially-observed Markov process (PanelPOMP) models. To do so, it relies on and extends a number of facilities that the pomp package provides for inference on time series data using partially-observed Markov process (POMP) models.

The panelPomp package extends to panel data some of the capabilities of the pomp package to fit nonlinear, non-Gaussian dynamic models. This is done accomodating both fixed and random effects. Currently, the focus is on likelihood-based approaches. In addition to these likelihood-based tools, panelPomp also provides a framework under which alternative statistical methods for PanelPOMP models can be developed (very much like pomp provides a platform upon which statistical inference methods for POMP models can be implemented).

Data analysis using panelPomp

The first step in using panelPomp is to encode one's model(s) and data in objects of class panelPomp. One does this via a call to the panelPomp constructor function.

panelPomp version 1.1 provides algorithms for

1. particle filtering of panel data (AKA sequential Monte Carlo or sequential importance sampling), as proposed in Bretó, Ionides and King (2020). This reference provides the fundamental theoretical support for the averaging of Monte Carlo replicates of panel unit likelihoods as implemented in panelPomp; see pfilter

2. the panel iterated filtering method of Bretó, Ionides and King (2020). This reference provides the fundamental theoretical support for the extensions of the iterated filtering ideas of Ionides et al. (2006, 2011, 2015) to panel data as implemented in panelPomp; see mif2

The package also provides various tools for handling and extracting information on models and data.

Extending the pomp platform for developing inference tools

panelPomp extends to panel data the general interface to the components of POMP models provided by pomp. In doing so, it contributes to the goal of the pomp project of facilitating the development of new algorithms in an environment where they can be tested and compared on a growing body of models and datasets.

Comments, bug reports, and requests

Contributions are welcome, as are suggestions for improvement, feature requests, and bug reports. Please submit these via the panelPomp issues page. We particularly welcome minimal working examples displaying uninformative, misleading or inacurate error messages. We also welcome suggestions for clarifying obscure passages in the documentation. Help requests are welcome, but please consider before sending requests whether they are regarding the use of panelPomp or that of pomp. For help with pomp, please visit pomp's FAQ.

Documentation

Examples are provided via the contacts(), panelGompertz() and panelRandomWalk() functions.

License

panelPomp is provided under the GPL-3 License.

Author(s)

Carles Bretó

References

Bretó, C., Ionides, E. L. and King, A. A. (2020) Panel Data Analysis via Mechanistic Models. Journal of the American Statistical Association, 115(531), 1178–1188. doi:10.1080/01621459.2019.1604367

See Also

pomp package, panelPomp