Netify Workshop

Package Overview

This vignette provides a high level overview of our package from start to finish. The best use of this vignette is to introduce the main components of the package to larger audiences. This overview covers our primary functions with data examples and minimal writing.

netify goals:

1. Create: Netify your data! We Make it easy for users to create networks from raw data in \(R\) as well as add additional features, such as nodal and dyadic variables, to the network object.
2. Explore: Explore characteristics of the network you created, like summary statistics at both the network and actor levels. Visualize your network.
3. Advance: Advance your network analysis to the next stage by preparing it for use in other network packages and modeling approaches.

netify provides a suite of primary functions to help achieve these goals:

netify can take dyadic data or edgelists to get started.

• The package can also create different types of networks including:

  • cross sectional networks
  • longitudinal (with static and varying actor composition)
  • bipartite networks
  • multilayer

• As well as create networks with different edge types:

  • weighted
  • binary
  • symmetric or non-symmetric

Step 1: Create 💡

Begin by loading packages and supplying the data. We will use the peacesciencer package to grab some familiar data.

# load packages
library(netify)

# install extra packages for this vignette
if(!'tidyverse' %in% rownames(installed.packages())){
  install.packages('tidyverse', repos='https://cloud.r-project.org') }
if(!'peacesciencer' %in% rownames(installed.packages())){
  install.packages('peacesciencer', repos='https://cloud.r-project.org') }
# load necessary packages for this vignette
library(peacesciencer)
library(tidyverse)

# organize external data for peacesciencer
peacesciencer::download_extdata()

# create dyadic data set over time using peacesciencer
cow_dyads <- create_dyadyears( 
     subset_years = c(1995:2014)
     ) |>
    # add mids
    add_cow_mids() |>
    # add capital distance
    add_capital_distance() |>
    # add democracy
    add_democracy() |>
    # add gdp
    add_sdp_gdp()

Next, create a netlet object from the above COW data frame using our package’s core function netify. There are a number of useful parameters, but the most important ones to highlight are:

• dyad_data is a dyadic data.frame that should have at least the following variables used to specify actors:

  • actor1: character indicating actor 1 variable in the data
  • actor2: character indicating actor 2 variable in the data

• netify_type is a type of netlet object (cross-sec, longit_list, or longit_array).

mid_long_network <- netify(
  cow_dyads,
  actor1='ccode1', actor2='ccode2', time='year',
  weight='cowmidonset', 
  actor_time_uniform=FALSE,
  sum_dyads=FALSE, symmetric=TRUE,
  diag_to_NA=TRUE, missing_to_zero=TRUE,
  nodal_vars = c('v2x_polyarchy1', 'v2x_polyarchy2'),
  dyad_vars = c('capdist'),
  dyad_vars_symmetric = c(TRUE)
  )

## ! Warning: Converting `actor1` and/or `actor2` to character vector(s).

Congratulations you have created a network object! 🎉

We can also add nodal and dyadic data after we’ve created the network via the add_nodal() and add_dyad() functions.

Let’s assume that we had information about each actor in the network that we’d like to add as a nodal variable after we already made the network object. This could be from our original data set or elsewhere. For example, lets add a logged variable measuring gdp for each node in the network over time:

# create a vector of nodal data
node_data <- unique(cow_dyads[,c("ccode1", "year", "wbgdppc2011est2")])
node_data$wbgdppc2011est2_log <- log(node_data$wbgdppc2011est2)

# add nodal variable to netlet object
mid_long_network <- add_nodal(
    netlet = mid_long_network,
    node_data = node_data,
    actor = "ccode1",
    time = "year"
)

# create another dyadic var in cow
cow_dyads$log_capdist = log(cow_dyads$capdist + 1)

# now lets add this to the netlet
mid_long_network <- add_dyad(
  netlet = mid_long_network, 
  dyad_data= cow_dyads,
  actor1= 'ccode1',
  actor2='ccode2',
  time='year',
  dyad_vars = 'log_capdist',
  dyad_vars_symmetric = TRUE
)

Step 2: Explore 🔎

We made a network, so let’s look at it. First, we might want to take a peek at the network object to see if the matrix looks the way we’d expect it to look. This function lets you glance at a specific slice of the network if it is longitudinal or the entire network if it is cross-sectional.

peek(
  mid_long_network, 
  when_to_peek = c('2012'), 
  what_to_peek = c(13, 15:24)
  )

## $`2012`
##      2 200 205 210 211 212 220 221 223 225 230
## 2   NA   0   0   0   0   0   0   0   0   0   0
## 200  0  NA   0   0   0   0   0   0   0   0   1
## 205  0   0  NA   0   0   0   0   0   0   0   0
## 210  0   0   0  NA   0   0   0   0   0   0   0
## 211  0   0   0   0  NA   0   0   0   0   0   0
## 212  0   0   0   0   0  NA   0   0   0   0   0
## 220  0   0   0   0   0   0  NA   0   0   0   0
## 221  0   0   0   0   0   0   0  NA   0   0   0
## 223  0   0   0   0   0   0   0   0  NA   0   0
## 225  0   0   0   0   0   0   0   0   0  NA   0
## 230  0   1   0   0   0   0   0   0   0   0  NA

Next, let’s examine a few basic summary statistics about the network using oursummary() function.

# create data.frame that provides network-level summary stats 
# for each year of the network
mid_long_summary <- summary(mid_long_network)

We can also make a quick visualization of network statistics over time using the summary statistics data frame.

plot_graph_stats(mid_long_summary)

We might also want to look at a summary of actor-level statistics overtime. Our built-in function, summary_actor will calculate in-degree, out-degree, average degree, and eigenvector centrality for each actor in each time period. The across_actors function allows users to toggle whether they want a summary of a given statistic across all actors (shown in a density plot) or for specific actors:

# every year & every actor 
summary_actor_mids <- summary_actor(mid_long_network)
head(summary_actor_mids)

##   actor time degree   prop_ties network_share closeness betweenness
## 1   100 1995      1 0.005376344    0.01470588         1           0
## 2   101 1995      1 0.005376344    0.01470588         1           0
## 3   110 1995      0 0.000000000    0.00000000       NaN           0
## 4   115 1995      0 0.000000000    0.00000000       NaN           0
## 5   130 1995      1 0.005376344    0.01470588         1           0
## 6   135 1995      1 0.005376344    0.01470588         1           0
##   eigen_vector
## 1            0
## 2            0
## 3            0
## 4            0
## 5            0
## 6            0

We can look at the distribution of the statistic for all actors over time:

# density plot across all actors
# for each stat
plot_actor_stats(
  summary_actor_mids,
  across_actor= TRUE,
  )