construct_graph: Construct directed acyclic graphs

In lawlerem/staRVe: Spatio-Temporal Analysis of Research surVEy data

Construct directed acyclic graphs

Description

Construct directed acyclic graphs

Usage

construct_persistent_graph(x, settings = new("settings"), silent = TRUE)

construct_persistent_graph_from_mesh(mesh, settings = new("settings"))

construct_transient_graph(
  x,
  y,
  time = 0,
  settings = new("settings"),
  silent = TRUE
)

construct_prediction_graph(pred, model, silent = TRUE)

Arguments

x, y	sf objects of point geometries
settings	An object of class settings
silent	Should intermediate calculations be shown?
mesh	An INLA mesh
time	A numeric vector containing the time index of each row in x.
pred	A long_stars object
model	A starve object

Value

An object of class dag; construct_dag instead returns a list with elements "locations" giving the sorted copy of locations and "dag" giving the constructed graph of class dag.

Functions

• construct_persistent_graph(): Construct a directed acyclic graph for a single sf object. This function uses a greedy algorithm to sort the locations of x. The first location of x is the first location of the sorted copy. Subsequent locations of the sorted copy are found iteratively by takingthe as-of-yet unsorted location closest to any of the already sorted locations.

  After sorting the locations, the directed acyclic graph is constructed by first creating an initial graph element with no "from" vertices and "to" vertices consisting of the first n sorted locations (where n equals the n_neighbours element of the settings argument). Any locations not present in the "to" vertices of the initial graph elements are given their own graph element with a single "to" vertex for that location and "from" vertices the n closest (by distance) locations before it in the sorted copy.

  A vertex with index i in either the "to" or "from" vertices represents the location in row i of the sorted copy of x.

• construct_persistent_graph_from_mesh(): Convert an INLA mesh to a persistent graph.

• construct_transient_graph(): Construct a directed acyclic graph with "to" vertices representing locations in x and "from" vertices representing locations in y. Unlike strv_construct_persistent_graph, neither x nor y are sorted, but the time argument must already be sorted when given as an argument (make sure that the locations in x are likewise sorted according7 to their respective times).

  The locations in x are first split into groups according to the values of thetime argument. Then within each group each location is given a graph element with a single "to" vertex for that location (in x) and whose "from" vertices are the n closest locations (in y) (where n equals the n_neighbours element of the settings argument).

  A vertex with index i in the "to" vertices represents the location in row i in the subset of x in the same time index group. Note that the same value of i may be used if there is more than one unique value in the time argument. A vertex with i in the "from" vertices represents the location in row i of y.

• construct_prediction_graph(): Works essentially the same as construct_transient_graph, with a few minor differences. Instead of directly supplying the locations for the "from" vertices, the "from" vertices are taken from the locations used in the persistent graph and transient graph from the model argument.

  The locations in pred are grouped according to their time index in pred. Each location is given a graph element with a single "to" vertex representing that location in pred. If that location is the same as any model location then the graph element has a single "from" vertex representing that model location, otherwise "from" vertices are the n closest model locations (where n equals the n_neighbours element of the settings argument).

  A vertex with index i in the "to" vertices represents the location in row i of pred (regardless of time index group). A vertex with index j in the "from" vertices represents the j'th location of the persistent graph if j <= k where k is the number of persistent graph locations, or represents the (j-k)'th location of the transient graph with the same time index as the "to" vertex.

See Also

dag_class For class definition of directed acyclic graphs

lawlerem/staRVe documentation built on Oct. 9, 2024, 4:48 a.m.