Step 1: Create data objects
Step 1a: Define node set(s) and attributes
We’ve loaded a dataset that defines its nodes and their attributes as a data frame. Let’s check what we have first.
class(actors)
#> [1] "data.frame"
head(actors)
#> label present floor gradeType
#> 1 Actor 1 TRUE 3 5
#> 2 Actor 2 TRUE 5 1
#> 3 Actor 3 TRUE 5 2
#> 4 Actor 4 TRUE 2 2
#> 5 Actor 5 TRUE 4 1
#> 6 Actor 6 TRUE 4 2Note that there are four column variables: label the
identifier, present, the floor of residence,
and gradeType their educational level (1 = freshmen to 5 =
graduate). All of these columns will be recognized as individual
attributes by goldfish.
We need to define them as nodes so that goldfish knows
what to do with them.
actors <- defineNodes(actors)
actors
#> Number of nodes: 84
#> Number of present nodes: 84
#>
#> First 6 rows
#> label present floor gradeType
#> 1 Actor 1 TRUE 3 5
#> 2 Actor 2 TRUE 5 1
#> 3 Actor 3 TRUE 5 2
#> 4 Actor 4 TRUE 2 2
#> 5 Actor 5 TRUE 4 1
#> 6 Actor 6 TRUE 4 2As you can see, the structure is the same, so we can still treat it like a data frame, but the added class helps goldfish interpret the data frame correctly.
Step 1b: Define networks
Next we want to define the dyadic or network elements: calls between our actors.
head(calls)
#> time sender receiver increment
#> 1 1220733470 Actor 72 Actor 50 1
#> 2 1221102974 Actor 43 Actor 51 1
#> 3 1221784293 Actor 43 Actor 51 1
#> 4 1221785882 Actor 43 Actor 22 1
#> 5 1221787264 Actor 43 Actor 55 1
#> 6 1221848443 Actor 43 Actor 51 1Note there are columns for time, sender,
and receiver. increment is a reserved
column.
To tell goldfish this is a network, we must define it as such:
?defineNetworkcallNetwork <- defineNetwork(nodes = actors, directed = TRUE)The argument directed is TRUE by default,
but we need to specify the nodes so that goldfish can check
for consistency and relate it to that nodeset as needed.
callNetwork
#> Dimensions: 84 84
#> Number of ties (no weighted): 0
#> Nodes set(s): actors
#> It is a one-mode and directed network
#>
#> First 6 rows and columns
#> receiver
#> sender Actor 1 Actor 2 Actor 3 Actor 4 Actor 5 Actor 6
#> Actor 1 0 0 0 0 0 0
#> Actor 2 0 0 0 0 0 0
#> Actor 3 0 0 0 0 0 0
#> Actor 4 0 0 0 0 0 0
#> Actor 5 0 0 0 0 0 0
#> Actor 6 0 0 0 0 0 0Note that we have not added any network data yet. By default,
defineNetwork() just constructs an empty matrix with
dimensions defined by the length of the nodeset(s). So we have an empty
network as a starting state.
Now that goldfish recognizes the matrix as a network, we can also
associate an event list that updates it. To do this we use the
linkEvents() function, which requires us to identify a
goldfish object to be updated, the events that update it and, in this
case, also the nodes that the events should relate to.
goldfish checks the consistency of all this information and
relates these objects to one another so that information can be called
as needed.
?linkEventscallNetwork <- linkEvents(x = callNetwork, changeEvent = calls, nodes = actors)
callNetwork
#> Dimensions: 84 84
#> Number of ties (no weighted): 0
#> Nodes set(s): actors
#> It is a one-mode and directed network
#> Linked events: calls
#>
#> First 6 rows and columns
#> receiver
#> sender Actor 1 Actor 2 Actor 3 Actor 4 Actor 5 Actor 6
#> Actor 1 0 0 0 0 0 0
#> Actor 2 0 0 0 0 0 0
#> Actor 3 0 0 0 0 0 0
#> Actor 4 0 0 0 0 0 0
#> Actor 5 0 0 0 0 0 0
#> Actor 6 0 0 0 0 0 0Task
You should now be able to do the same with the friendship nomination
network/event list. See the familiar columns time,
sender, and receiver. The new column,
replace, is an alternative treatment to
increment. When an event occurs, goldfish will replace the
value in the relevant cell with the value in this column instead of
incrementing it. Friendship is thus a binary network.
head(friendship)
#> time sender receiver replace
#> 1 1220918400 Actor 47 Actor 2 1
#> 3 1220918400 Actor 57 Actor 2 1
#> 4 1220918400 Actor 9 Actor 38 1
#> 5 1220918400 Actor 68 Actor 40 1
#> 6 1220918400 Actor 23 Actor 40 1
#> 7 1220918400 Actor 49 Actor 40 1
friendshipNetwork <- defineNetwork(nodes = actors, directed = TRUE)
friendshipNetwork <- linkEvents(
x = friendshipNetwork,
changeEvents = friendship,
nodes = actors
)
friendshipNetwork
#> Dimensions: 84 84
#> Number of ties (no weighted): 0
#> Nodes set(s): actors
#> It is a one-mode and directed network
#> Linked events: friendship
#>
#> First 6 rows and columns
#> receiver
#> sender Actor 1 Actor 2 Actor 3 Actor 4 Actor 5 Actor 6
#> Actor 1 0 0 0 0 0 0
#> Actor 2 0 0 0 0 0 0
#> Actor 3 0 0 0 0 0 0
#> Actor 4 0 0 0 0 0 0
#> Actor 5 0 0 0 0 0 0
#> Actor 6 0 0 0 0 0 0Step 1c: 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 list and the node list.
?defineDependentEventscallsDependent <- defineDependentEvents(
events = calls, nodes = actors,
defaultNetwork = callNetwork
)
callsDependent
#> Number of events: 439
#> Nodes set(s): actors
#> Default network: callNetwork
#>
#> First 6 rows
#> time sender receiver increment
#> 1 1220733470 Actor 72 Actor 50 1
#> 2 1221102974 Actor 43 Actor 51 1
#> 3 1221784293 Actor 43 Actor 51 1
#> 4 1221785882 Actor 43 Actor 22 1
#> 5 1221787264 Actor 43 Actor 55 1
#> 6 1221848443 Actor 43 Actor 51 1Intermediate step: Visualization
While not a required part of the modeling process, we highly
recommend the visualization of your data for analytic and diagnostic
purposes. goldfish includes wrappers for base
R commands to help extract monadic and dyadic information
for certain time points, ?as.data.frame.nodes.goldfish and
?as.matrix.network.goldfish .
We can use these functions to visually compare our network at two (or
more) different time periods using migraph. See the
migraph package documentation for additional information
about network visualization.
library(ggraph)
library(migraph)
# The network at the beginning
callNetworkBgn <- as.matrix(callNetwork)
autographr(callNetworkBgn, labels = FALSE, layout = "fr")
plot of chunk plot-teaching1
# The network at half time
callNetworkHlf <- as.matrix(callNetwork,
time = calls$time[floor(nrow(calls) / 2)]) |>
add_node_attribute("floor", actors$floor)
autographr(callNetworkHlf, labels = FALSE, layout = "fr") +
geom_node_point(aes(color = as.factor(floor)), size = 2, show.legend = FALSE)
plot of chunk plot-teaching1
# The network at the end
callNetworkEnd <- as.matrix(callNetwork, time = max(calls$time) + 1) |>
add_node_attribute("floor", actors$floor)
autographr(callNetworkEnd, labels = FALSE, layout = "fr") +
geom_node_point(aes(color = as.factor(floor)), size = 2, show.legend = FALSE)
plot of chunk plot-teaching1
# The tie strength at the end
table(as.matrix(callNetwork, time = max(calls$time) + 1))
#>
#> 0 1 2 3 4 6 7 8 9 10 11 13 15 64 115
#> 6972 36 13 10 8 1 5 3 1 1 1 2 1 1 1