Statistical Network Models for Dynamic Network Data.
goldfish 
Description
The {goldfish} package offers a collection of tools designed for applying statistical models to dynamic network data. It primarily focus on models for relational event data, namely, sequences of interactions between actors or entities within a network, enriched by fine-grained time-stamps information. Relational event data emerge in various domains, such as automatically collected data about interactions in communication and social media research, social science studies using social sensors, and archival network studies that provide in-depth details regarding the timing or sequence of relational actions between nodes.
Currently, the package includes the following models:
- Dynamic Network Actor Models (DyNAM): Investigate relational event models as an actor-oriented decision process (Stadtfeld and Block 2017).
- rate: Actors compete for creating the next relational event.
- choice: The active actor choose the receiver of the event.
- choice_coordination: The creation of coordination ties as a two-sided process (Stadtfeld, Hollway, and Block 2017)
- Dynamic Network Actor Models for interactions (DyNAMi): Investigate dynamics of conversation groups and interpersonal interaction in different social contexts from an actor-oriented perspective (Hoffman et al. 2020).
- rate: Actors compete for joining or leaving groups.
- choice: The active actor choose the group to join.
- Relational Event Models (REM): Investigate relational event models as a tie-oriented process (Butts 2008).
Vignettes
For detailed documentation on each model, including usage examples, users are encouraged to consult the package’s vignettes and help files:
- Getting Started with goldfish (DyNAM and REM)
- Coordination ties (DyNAM-choice coordination)
- Face to face interactions (DyNAMi)
- Catalog of available effects
Table of Contents
Installation
You can install {goldfish} directly from CRAN:
install.packages("goldfish")
To install the development version from GitHub, use the remotes package:
- For latest stable version:
remotes::install_github("stocnet/goldfish", build_vignettes = TRUE) - For latest development version:
remotes::install_github("stocnet/goldfish@develop", build_vignettes = TRUE)
Or by downloading and install the latest binary releases for all major OSes – Windows, Mac, and Linux – can be found here.
Installing OpenMP on Mac OSX
In some cases, you may get an error that does not allow installation of {goldfish} from source on Mac OSX versions, including under R 4.0.0. The error may relate to compiling the parts of {goldfish} that are written in C++, or whether OpenMP (for parallelisation) can be found.
Many installation woes can be solved by directing R to use Homebrew installed gcc. An updated setting up instructions thanks to @timonelmer are available here.
More details can be found here (Thank you @Knieps for identifying this.). Other links that may be helpful include:
- https://asieira.github.io/using-openmp-with-r-packages-in-os-x.html
- https://thecoatlessprofessor.com/programming/cpp/r-compiler-tools-for-rcpp-on-macos/
- https://ryanhomer.github.io/posts/build-openmp-macos-catalina-complete
- https://pat-s.me/transitioning-from-x86-to-arm64-on-macos-experiences-of-an-r-user/
Please share feedback on which of these work and we will update the installation guide accordingly.
Usage
Below is a quick-start guide to using the {goldfish} package. The dataset used in this example is an abbreviated version of the MIT Social Evolution data (?Social_Evolution).
Define data objects and link events
The main data objects required for the analysis are the node set(s) defineNodes() and network(s) defineNetwork(). The node set object contains labels and attributes of the actors in the network. In contrast, a network object contains the information of past relational events between actors. By default, defineNetwork() constructs an empty matrix, its dimensions defined by the length of the nodeset(s). Data frames containing event data that modify these data objects can be linked to them using the linkEvents() method.
library(goldfish)
data("Social_Evolution")
callNetwork <- defineNetwork(nodes = actors, directed = TRUE) |> # 1
linkEvents(changeEvent = calls, nodes = actors) # 2
The events data frame, which indicates the time-varying attributes in the node set, contains the following columns:
time: The time when the attribute changes, either anumericorPOSIXctvalue.node: The node for which the attribute changes, acharactervalue that matches thelabelvariable in the node set.replace: The new value of the attribute, anumericvalue.
The events data frame that details the relational events between actors contains the following columns:
time: The time when the event occurred, either anumericorPOSIXctvalue.sender: The actor initiating the event, acharactervalue that matches thelabelvariable in the node set.receiver: The actor receiving the event, acharactervalue that matches thelabelvariable in the node set.incrementorreplace: Anumericvalue indicating either the increment that the relational event represents or the new value.
Define dependent events
The final step in defining the data objects is to identify the dependent events. Here we would like to model as the dependent variable the calls between individuals. We specify the event data frame and the node set.
callsDependent <- defineDependentEvents(
events = calls, nodes = actors,
defaultNetwork = callNetwork
)
Model specification and estimation
We specify our model using the standard R formula format like:
goldfish_dependent ~ effects(process_state_element)
We can see which effects are currently available and how to specify them here:
vignette("goldfishEffects")
Now to estimate this model, we use the ?estimate function.
mod00Rate <- estimate(
callsDependent ~ indeg + outdeg,
model = "DyNAM", subModel = "rate"
)
summary(mod00Rate)
#>
#> Call:
#> estimate(x = callsDependent ~ indeg + outdeg, model = "DyNAM",
#> subModel = "rate")
#>
#>
#> Coefficients:
#> Estimate Std. Error z-value Pr(>|z|)
#> indeg 0.551445 0.066344 8.3119 < 2.2e-16 ***
#> outdeg 0.263784 0.028386 9.2927 < 2.2e-16 ***
#> ---
#> Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
#>
#> Converged with max abs. score of 2e-05
#> Log-Likelihood: -1750.9
#> AIC: 3505.8
#> AICc: 3505.9
#> BIC: 3514
#> model: "DyNAM" subModel: "rate"
mod00Choice <- estimate(
callsDependent ~ inertia + recip + trans,
model = "DyNAM", subModel = "choice"
)
summary(mod00Choice)
#>
#> Call:
#> estimate(x = callsDependent ~ inertia + recip + trans, model = "DyNAM",
#> subModel = "choice")
#>
#>
#> Coefficients:
#> Estimate Std. Error z-value Pr(>|z|)
#> inertia 5.19690 0.17397 29.8725 < 2.2e-16 ***
#> recip 1.39802 0.17300 8.0812 6.661e-16 ***
#> trans -0.23036 0.21554 -1.0687 0.2852
#> ---
#> Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
#>
#> Converged with max abs. score of 7e-05
#> Log-Likelihood: -696.72
#> AIC: 1399.4
#> AICc: 1399.5
#> BIC: 1411.7
#> model: "DyNAM" subModel: "choice"
About
This project is a joint collaboration between the Social Networks Lab at ETH Zürich and the Graduate Institute Geneva, and incorporates and supports several sub-projects.
References
Butts, C. T. 2008. “A Relational Event Framework for Social Action.” Sociological Methodology 38 (1): 155–200.
Hoffman, Marion, Per Block, Timon Elmer, and Christoph Stadtfeld. 2020. “A Model for the Dynamics of Face-to-Face Interactions in Social Groups.” Network Science 8 (S1): S4–25. https://doi.org/10.1017/nws.2020.3.
Stadtfeld, C., and P. Block. 2017. “Interactions, Actors, and Time: Dynamic Network Actor Models for Relational Events.” Sociological Science 4 (14): 318–52. https://doi.org/10.15195/v4.a14.
Stadtfeld, C., J. Hollway, and P. Block. 2017. “Dynamic Network Actor Models: Investigating Coordination Ties Through Time.” Sociological Methodology 47 (1): 1–40. https://doi.org/10.1177/0081175017709295.