Network metrics
network_metrics.RmdNetwork metrics can be extracted using the function
extract_metric(bison_model, metric_name). The
metric_name variable is made up of two parts: the network
feature and the network metric. For example, node_eigen
will calculate the eigenvector centrality of each node.
The following feature/metric pairs are currently supported in bisonR:
-
edge- Edge-level properties-
weight- Edge weight
-
-
node- Node-level properties-
degree[t]- Node degree (where t is the binary threshold for an edge) -
strength- Node strength -
eigen- Node eigenvector centrality -
betweenness- Node betweenness -
closeness- Node closeness -
clustering- Node-level clustering coefficient
-
-
global- Network-level properties-
density- Weighted density -
cv- Coefficient of variation of edge weights - also known as social differentiation -
std- Standard deviation of edge weights -
diameter- Diameter of the network -
clustering- Clustering coefficient of the network - also known as transitivity
-
Examples
See below for examples of how to generate the network metrics listed above.
library(bisonR)
#> Loading required package: cmdstanr
#> This is cmdstanr version 0.9.0
#> - CmdStanR documentation and vignettes: mc-stan.org/cmdstanr
#> - CmdStan path: /home/runner/.cmdstan/cmdstan-2.36.0
#> - CmdStan version: 2.36.0
library(dplyr)
#>
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#>
#> filter, lag
#> The following objects are masked from 'package:base':
#>
#> intersect, setdiff, setequal, union
sim_data <- simulate_bison_model("binary", aggregated = TRUE)
df <- sim_data$df_sim
fit_edge <- bison_model(
(event | duration) ~ dyad(node_1_id, node_2_id),
data=df,
model_type="binary_conjugate",
priors=get_default_priors("binary_conjugate")
)Edge
Edge weight
metrics <- extract_metric(fit_edge, "edge_weight", num_draws=10)
metrics[1:6, 1:5] # Preview first few samples and edges
#> [,1] [,2] [,3] [,4] [,5]
#> [1,] -1.7763303 1.2071965 1.3689566 3.2731301 3.4882468
#> [2,] 0.9357479 0.9649296 1.0682155 1.1780742 0.8553972
#> [3,] -3.6258711 1.8834824 2.0402320 0.2606149 2.6947782
#> [4,] -2.7593545 3.0038231 1.1405372 1.8913551 5.6115489
#> [5,] -1.8913159 0.9642772 0.8170626 4.0066255 1.8581266
#> [6,] 0.1826523 1.2889817 1.4426996 0.8583163 2.2183483Node
Node degree
metrics <- extract_metric(fit_edge, "node_degree[0.2]")
metrics[1:6, 1:5] # Preview first few samples and nodes
#> [,1] [,2] [,3] [,4] [,5]
#> [1,] 8 7 5 6 8
#> [2,] 8 7 8 8 4
#> [3,] 8 5 7 8 7
#> [4,] 9 6 5 8 8
#> [5,] 7 6 8 7 6
#> [6,] 8 5 8 7 6Node strength
metrics <- extract_metric(fit_edge, "node_strength", num_draws=10)
metrics[1:6, 1:5] # Preview first few samples and nodes
#> [,1] [,2] [,3] [,4] [,5]
#> [1,] 5.250600 3.605330 4.044063 4.328342 4.147407
#> [2,] 6.390623 3.869768 4.278853 5.145356 4.451589
#> [3,] 6.153401 3.094590 4.164048 4.940242 4.975292
#> [4,] 6.359865 3.531231 5.004913 5.160148 4.461221
#> [5,] 5.756916 3.451609 4.778303 4.609269 4.311387
#> [6,] 5.174642 3.617823 5.412316 4.863402 3.786787Node eigenvector centrality
metrics <- extract_metric(fit_edge, "node_eigen")
metrics[1:6, 1:5] # Preview first few samples and nodes
#> [,1] [,2] [,3] [,4] [,5]
#> [1,] 1 0.6998075 0.8274551 0.7619760 0.7679569
#> [2,] 1 0.5417456 0.8373102 0.8503530 0.8265666
#> [3,] 1 0.6760519 0.8898821 0.9472859 0.9077447
#> [4,] 1 0.4732591 0.7720060 0.8812934 0.7572672
#> [5,] 1 0.4138459 0.8436132 0.9317520 0.8012959
#> [6,] 1 0.5517734 0.8733358 0.9190946 0.6936717Node betweenness
metrics <- extract_metric(fit_edge, "node_betweenness")
metrics[1:6, 1:5] # Preview first few samples and nodes
#> [,1] [,2] [,3] [,4] [,5]
#> [1,] 4 0 0 4 1
#> [2,] 9 0 0 1 6
#> [3,] 7 0 0 6 0
#> [4,] 14 0 0 3 6
#> [5,] 12 0 3 1 2
#> [6,] 8 2 5 0 3Node closeness
metrics <- extract_metric(fit_edge, "node_closeness")
metrics[1:6, 1:5] # Preview first few samples and nodes
#> [,1] [,2] [,3] [,4] [,5]
#> [1,] 0.06313667 0.03415140 0.05568646 0.05583207 0.04863124
#> [2,] 0.05638048 0.03776727 0.05059571 0.05298703 0.05238034
#> [3,] 0.06271996 0.03884989 0.05684554 0.05761289 0.05130179
#> [4,] 0.05566669 0.02823852 0.05793152 0.04768573 0.04804053
#> [5,] 0.05791287 0.05182419 0.05418053 0.06348239 0.05845334
#> [6,] 0.06342691 0.03999048 0.05167693 0.05367846 0.04884259Node-level clustering coefficient
metrics <- extract_metric(fit_edge, "node_clustering[0.2]")
metrics[1:6, 1:5] # Preview first few samples and nodes
#> [,1] [,2] [,3] [,4] [,5]
#> [1,] 0.6111111 0.8333333 0.6666667 0.6666667 0.8000000
#> [2,] 0.7857143 0.7142857 0.7619048 0.9047619 0.9333333
#> [3,] 0.7142857 0.6071429 0.7333333 0.7333333 0.6666667
#> [4,] 0.7142857 0.6000000 0.6190476 0.6666667 0.8000000
#> [5,] 0.6428571 0.7000000 0.6666667 0.6785714 0.6666667
#> [6,] 0.5833333 0.8000000 0.7619048 0.5714286 0.9000000Global
Global coefficient of variation
metrics <- extract_metric(fit_edge, "global_cv")
metrics[1:6] # Preview first few samples
#> [1] 0.6620834 0.7752250 0.7524315 0.6514228 0.6969299 0.8203320Global density
metrics <- extract_metric(fit_edge, "global_density")
metrics[1:6] # Preview first few samples
#> [1] 0.3930515 0.4329158 0.4216413 0.4753094 0.4301271 0.4341689Global standard deviation of edge weights
metrics <- extract_metric(fit_edge, "global_std")
metrics[1:6] # Preview first few samples
#> [1] 0.3093545 0.3094215 0.3027798 0.3243845 0.3032741 0.3233578Global diameter
metrics <- extract_metric(fit_edge, "global_diameter")
metrics[1:6] # Preview first few samples
#> [1] 0.3318503 0.2553142 0.4628383 0.2985160 0.4401158 0.5826086Global clustering coefficient
metrics <- extract_metric(fit_edge, "global_clustering[0.2]")
metrics[1:6] # Preview first few samples
#> [1] 0.5827338 0.7288136 0.6580645 0.7853403 0.8000000 0.8111588