NEWS.md
network.goldfish objects (and linked events and nodelists) to migraph-compatible objects (closed #96)add_node_attributes() to add_node_attribute() and add_edge_attributes() to add_tie_attribute()
is_*() are now considered graph-level ‘marks’node_is_isolate() for marking isolatesnode_is_max(), node_is_min(), tie_is_max(), tie_is_min() for converting ‘measures’ into ‘marks’graph_motif (fixed #234)cluster_concor() and cluster_hierarchical() are now exportedautographr() no longer requires “highlight_measure” and “identify_function” arguments as users can now convert ‘measures’ to ‘marks’ and use these for “node_color” or “edge_color”@format for more consistent documentationison_brandes2 dataset, a two-mode version of the original one-mode datasetmpn_cow_trade and mpn_cow_igo datasets (thanks @JaelTan)mpn_elite_mex
node_reach() for calculating reach centrality (closed #196){oaqc} are now Suggests; the user is prompted to install them when required in autographr(), plot.member(), and node_quad_census() respectively@family tags for improved cross-referencingcreate_ and generate_ functions now:n passed an existing networkcreate_tree(), generate_smallworld(), generate_scalefree()
create_ functions can now take a membership vector or split into equal partitions by defaultcreate_components() no longer accepts a number of components, instead relying on the membership vectorcreate_core() for creating core-periphery graphsgenerate_random() now inherits attributes from any networkto_ functions useful for working with networks of different typesto_redirected() for adding or swapping direction to networks (closed #219)to_blocks() for reducing a network down by a membership vector; blockmodel() and reduce_graph() are now deprecatedto_multilevel.igraph() now only works on two-mode networks; returns the original network if passed a one-mode networkis_ functionsis_signed.data.frame() and is_signed.matrix() now rely on new helper is.wholenumber() rather than misleading is.integer()
is_directed.igraph() and is_directed.matrix() now return FALSE for two-mode networksis_connected() now returns result for strong components if directed and weak components if undirectedas_igraph.data.frame() now infers third column as weightedge_multiple(), edge_loop(), edge_reciprocal() moved from measuresedge_bridges()
"edge_measure" S3 class has been added, along with print() and plot() methodssummary.node_measure() method for printing a summary by a membership vector; summarise_statistics() is now deprecated"graph_measure" class resultsgraph_components() now calculates strong components for directed networks else weak componentsprint.graph_measure() now correctly labels two-mode results where a vector is givengraph_smallworld()graph_core() for calculating correlation of an observed network to a core-periphery network of the same dimensions (closed #39)graph_factions() for calculating correlation of an observed network to a component network of the same dimensions (closed #40)graph_modularity() for calculating modularity of an observed network, including modularity for two-mode networks (closed #144)node_constraint()node_bridges(), node_redundancy(), node_effsize(), node_efficiency(), node_hierarchy()
node_betweenness() no longer needs nobigint argument; just uses default from igraph
"node_motif" S3 class for the output of node_*_census() functionsprint.node_motif() for tibble-printing of census resultssummary.node_motif() to summarise censuses by a membership vector, replacing group_tie_census() and group_triad_census(), which are now deprecated"graph_motif" S3 class for the output of graph_*_census() functionsnode_path_census() for returning the shortest distances from each node to every other node (closed #222)node_tie_census() now creates unique column names"node_member" S3 class for vectors of nodes’ cluster membershipsplot.node_member() replaces ggtree()
node_equivalence() for identifying nodes’ membership in classes equivalent with respect to some arbitrary motif census"hierarchical" and "concor" now options for cluster within node_equivalence(); blockmodel_concor() is now deprecated (closed #123)"elbow" now an option for k selection within node_equivalence(); ggidentify_clusters() is now deprecated"silhouette" and "strict" options for k selection (closed #197)k to be defineddistance passed to stats::dist that defines the distance metric (closed #36)range constrains the number of k evaluated by "elbow" and "silhouette" to improve parsimony and avoid long elapsed timesnode_automorphic_equivalence() for identifying nodes’ membership in automorphically-equivalent classes (closed #187)node_structural_equivalence() replaces cluster_structural_equivalence()
node_regular_equivalence() replaces cluster_regular_equivalence()
node_coreness() for nodes’ k-core score (closed #200)node_strong_components() and node_weak_components() for more direct calls; node_components() now calculates strong components for directed networks else weak components"graph_test" S3 class replaces "cug_test" and "qap_test"plot.graph_test() replaces plot.cug_test() and plot.qap_test()
print.graph_test() methodplot.matrix() now plots adjacency/incidence matrices with sorting and horizontal/vertical lines if a membership vector is provided, effectively replacing plot.block_model()
autographr() can highlight nodes that max (by default) some measure (thank you, @BBieri, closed #224)layout_tbl_graph_stressgrid() as an extra optionggatyear() is deprecated{roxytest}
m argument into p for generate_random(), p can now be passed an integer to indicate the number of ties the network should haveto_edges() to be ~26 times faster on averageto_edges()
to_subgraph() now instead of dplyr::filter() or strain()
"measure" class "node_measure" and added "graph_measure" class with print methodnormalized is now the second argumentdirected and weights arguments have been removed and are now imputed, if this is undesired please use to_*() firstnode_degree() now calculates strength centrality if network is weightednode_eigenvector() and graph_eigenvector() both work with two-mode networksedge_degree() and edge_eigenvector(), which both just apply the corresponding nodal measure to the edge graphedge_mutual() renamed to edge_reciprocal()
graph_assortativity()
directed and direction arguments in some functions; whereas directed is always logical (TRUE/FALSE), direction expects a character string, e.g. “in”, “out”, or “undirected”generate_permutation() now has an additional logical argument, with_attr, that indicates whether any attributes from the original data should be passed to the permuted objectcreate_*() functions now accept existing objects as their first argument and will create networks with the same dimensionsread_pajek() now imports nodal attributes alongside the main edgesread_ucinet() now enjoys clearer documentationas_*() functions now retain weights where present; if you want an unweighted result, use is_unweighted() afterwardsas_edgelist.network() now better handles edge weightsas_matrix.igraph() now better handles edge signsis_twomode(), is_directed(), is_weighted(), is_labelled(), is_signed(), is_multiplex(), is_complex(), and is_graph()
as_edgelist(), and to_unweighted(), and improved the data.frame method for as_matrix()
to_named() and to_unsigned()
to_edges() for creating adjacency matrices using a network’s edges as nodesproject_rows() and project_cols() functions to to_mode1() and to_mode2(), which is both more consistent with other functions naming conventions and more generic by avoiding the matrix-based row/column distinctionnode_mode(), which returns a vector of the mode assignments of the nodes in a networkedge_signs(), which returns a vector of the sign assignments of the edges in a networkautographr() works and how ggraph extends thisautographr() now incorporates ggidentify() functionality (closed #150)ggatyear()
measure class and directed most node_*() functions to create objects of this classggdistrib() and offers “hist” and “dens” methods for histograms and density plots respectivelyedge_betweenness() wraps igraph’s function of the same nameedge_closeness() measures the closeness centrality of nodes in an edge adjacencynode_cuts() identifies articulation points (nodes) in a networkedge_bridges() identifies edges that serve as bridges in a networkgraph_cohesion() measures how many nodes would need to be removed to increase the number of components (closed #192)graph_adhesion() measures how many edges would need to be removed to increase the number of componentsgraph_length() measures the average path lengthgraph_diameter() measures the longest path lengthnode_smallworld() and added graph_smallworld(), which works with both one- and two-mode networks (fixed #214)node_quad_census()
network_reg()’s formula-based systemnetwork_reg() can now handle binary and multiple categorical variables (factors and characters, closed #211);network_reg() can now manage interactions specified in the common syntax; var1 * var2 expands to var1 + var2 + var1:var2 (closed #163)dist() and sim() effects have been added (closed #207)network_reg() now employs logistic regression to estimate a binary outcome and linear regression to estimate a continuous outcome (closed #184)network_reg() now uses Dekker et al’s semi-partialling procedure by default for multivariate specifications (closed #206), defaulting to y-permutations in the case of a single predictor (closed #208)network_reg(), relying on furrr for potential parallelisation and progressr for progress reports (closed #185, #186)test_random() and test_permutation(), relying on furrr for potential parallelisation and progressr for progress reports; note that nSim argument now times (closed #199)netlm and netlogit class objects (closed #183)netlm and netlogit class objects (closed #216), which plots the empirical distribution for each test statistic, indicates percentiles relating to common critical values, and superimposes the observed coefficientscug_test and qap_test class objects, which plots the empirical distribution, highlighting tails beyond some critical value (closed #213), and superimposing the observed coefficient and, possibly, 0print.block_model() replaces print.blockmodel()
plot.block_model() replaces plot.blockmodel()
ison_southern_women instead of southern_women
ison_brandes instead of brandes
ison_networkers instead of ison_eies
ison_algebra instead of ison_m182
ison_adolescents instead of ison_coleman
mpn_elite_mex is extended with data from Pajek and with help from Frank Heberison_networkers becomes named with information from tnet
mpn_* and ison_* datasets, including references/sourcescluster_structural_equivalence() for isolatesadd_ functionsis_ methods: is_multiplex(), is_uniplex(), is_acyclic()
edge_ functions to identify edges by properties: edge_mutual(), edge_multiple(), edge_loop()
read_nodelist() and read_edgelist()
as_network() method to convert correctly form an igraph to a network object.ggraphgrid() documentationread_edgelist() and read_nodelist() to readxl to avoid Java dependencywrite_edgelist() and write_nodelist() to avoid Java dependencyas_network(), as_igraph(), and is_directed()
read_ and write_ functions and updated documentationread_edgelist() for importing edgelists from Excel and csv filesread_pajek() for importing .net and .paj fileswrite_edgelist(), write_nodelist, write_pajek(), and write_ucinet() for exporting into various file formats (Excel, csv, Pajek, and UCINET)is_graph() to check if an object is a graph or notas_network() to retain attributesas_ and to_ functionsas_ functions to convert from dataframes instead of tibblesas_igraph() functionto_undirected() function to work with network objectsto_main_component() function so that it retains vertex attributes in network objectsedge_attribute() to grab a named edge attribute from a graph/networkto_unweighted() to prevent conversion of network object into igraph object when deleting weightssummarise_statistics()
network_reg() exampleautographr()
as_matrix() method for networks now works with two-mode and weighted networksas_igraph() method for matrices now checks for weights independently of coercionas_igraph() method for networks now works with two-mode and weighted networksas_network() method for matrices now works with two-mode and weighted networksas_network() method for edgelists, igraph, and tidygraphs now works with weighted networksto_unnamed() method for edge liststo_simplex() method for matricesto_main_component() method for networksto_multilevel() method for matricesmutate_edges() now coalesces rows of edgesgraph_blau_index()
generate_permutation() and thus test_permutation()
netlm() to network_reg() to avoid frustrating conflictsnetwork_reg() now accepts migraph-consistent objectsnetwork_reg() now accepts formula terms such as ego(), alter(), and same()
generate_permutation() which takes an object and returns an object with the edges permuted, but retaining all nodal attributesgenerate_random() also work with an existing object as input, in which it will return a random graph with the same dimensions and densitygraph_blau_index()
graph_ei_index()
test_random() (defunct test_cug())autographr() arguments take variable names in quotation marksnode_names() for quickly accessing node labelsnode_attribute() for quickly accessing a certain nodal attributeedge_weights() for quickly accessing edge weightsgraph_nodes() for quickly accessing a count of nodes in the graph, note that for two-mode networks this will be a vector of length 2graph_edges() for quickly accessing a count of edges in the graphgraph_dimensions() is currently a copy of graph_nodes()
add_node_attributes() for adding particular nodal attributesadd_edge_attributes() for adding edges from another graphcopy_edge_attributes() for copying all nodal attributes from another graphgraph_blau_index() for summarising diversity of an attribute in a network or groupgraph_ei_index() for summarising diversity of an attribute over a network’s tiesnode_quad_census(), especially useful for two-mode blockmodellinggraph_mixed_census()
test_random() carries out a conditional uniform graph (CUG) testtest_permutation() carries out a quadratic assignment procedure (QAP) testaes() from ggplot2
node_shape option to autographr()
as_edgelist() methods for converting other objects into edgeliststo_unnamed() on ‘network’ objects now operates on them directlyto_ documentation significantlyto_onemode() that was tripping blockmodel() on networks that are already one-modeis_connected() to test whether network is connected, method = argument can be specified as weak or strong
create_tree() and create_lattice(), and made create_star() a bit faster for one-mode networksgenerate_smallworld() and generate_scalefree(), though only for one-mode networks currently=2
graph_dyad_census() for more graph profile optionsblockmodel_concor() when an object was of class ‘igraph’ but not ‘tbl_graph’blockmodel() was treating two-mode networks"elbow" and "strict" methods for k-identificationggidentify_clusters() for speedas_matrix() where availablegraph_equivalency() into the same for two-mode networks and graph_congruency()for three-mode (two two-mode) networksgraph_reciprocity() methodgraph_components() and node_components()
node_tie_census() output so that it’s consistent with node_triad_census() and future node_census functionscluster_triad_census() to group_triad_census()
group_tie_census()
autographr() for plotting convex/concave hullsggraphgrid() a set of layout functions:layout_tbl_graph_frgrid() or autographr(object, "frgrid") for snapping Fruchterman-Reingold to a gridlayout_tbl_graph_kkgrid() or autographr(object, "kkgrid") for snapping Kamada-Kawai to a gridlayout_tbl_graph_gogrid() or autographr(object, "gogrid") for snapping graph optimisation to a gridggraphgrid() has been deprecatedas_igraph()
to_uniplex() was not returning a weighted graphblockmodel() was not retaining node names in all parts of the object structureplot_releases() to another packageis_signed() to logically test whether the network is a signed networkto_unsigned() for extracting networks of either “positive” or “negative” ties from a signed networktbl_graph methods for all other to_ functionsactivate() from tidygraph
autographr()
plot_releases() from this packagegraph_balance() to be much faster, following David Schoch’s signnet package (see that package for further extensions)ison_marvel_relationships to be a double (-1/1) to be compatible with the new graph_balance() and signnet
to_main_component() to extract the main component of a networkto_onemode() for moving to multimodal igraph objectsto_uniplex() method to delete edge types and their edges from multiplex networksto_simplex() method to delete loops from a networkto_named() method for randomly naming unlabeled networksison_mm, ison_mb, ison_bm, and ison_bb projection illustration dataison_karateka community detection illustration dataison_marvel_teams and ison_marvel_relationships datasetsison_m182 dataset of friends, social and task ties between 16 anonymous studentsadolescent_society dataset to ison_colemanfor consistencygraph_eigenvector() for one mode networksgraph_balance() for measuring structural balancenode_tie_census(), node_triad_census(), cluster_triad_census(), and graph_triad_census()
graph_clustering() into the cohesion measures graph_density(), graph_reciprocity(), graph_transitivity(), and graph_equivalence()
node_smallworld() to use separated cohesion measuresblockmodel() which masks its sna namesake but has the advantages of working with two-mode networks and retaining node names where availablecluster_structural_equivalence() and cluster_regular_equivalence() as bases for blockmodellingreduce_graph() for creating a network from a blockmodelautographr() for plotting graphs with sensible defaultsggraphlabel() since core functionality now provided by autographrggidentify() to identify the node with the highest value of a specified node-level measureggatyear() for subsetting and plotting edgelists at yeargglineage() to return a graph colored according to lineageggtree() for neatly visualising hierarchical clustersggidentify_clusters() for identifying which number of clusters is most appropriate following the elbow methodggraph::theme_graph() present in a few different visualisation functionsbrandes dataset for teaching centrality measuresadolescent_society dataset for teaching friendship paradoxread_edgelist() for importing Excel-created edgelists directlyggraphlabel() for one-function (1F) plotting label-based network graphsggevolution() for 1F-plotting begin/end graph comparisonsggraphgrid() for 1F snap-to-grid graph layouts based on Fruchterman-Reingold or Kamada-Kawaiggidentify() for 1F identifying nodes with maximum scores based on some arbitrary functionto_undirected() for symmetrising networks of all typesto_ functions S3 methodsClosed #100 by converting as_ coercion functions to S3 methods
as_matrix() weightingas_tidygraph()
gglineage() for graphing a citation network through timeggevolution() for graphing two timepoints of the same network side by sideggraphgrid() for locking a graph to a gridplot.igraph() defaultsFixed bug in as_matrix() with frame matrix by dropping (rarely necessary) functionality
as_ functionsbinarise() to to_unweighted()
to_unnamed() for unlabelling networksbinarise() for unweighting networksas_tidygraph() when passed a tbl_graph directlyUpdated README
Added CITATION details
Separated coercion (previously conversion) and manipulation
Added some more inter-class coercion tests
Fixed bug in how as_network() sometimes coerced two-mode networks into much larger dimension matrices
Added more is_ tests for class-independent property tests
is_weighted()
is_directed()
is_labelled()
Added @csteglich ’s read_ucinet() and write_ucinet() functions
read_ucinet() offers a file-picker when file path unknownread_ucinet() now imports to an igraph-class object by default, with an argument to allow other alternativeswrite_ucinet() works with all migraph-compatible objectsUpdated mpn_bristol documentation
Added create_star() function
directed = argumentcreate_ documentationRenamed sample_affiliation() to generate_random()
generate_random() to be able to generate random one- or two-mode networks2021-04-11
as_network() to coerce objects into network class2021-03-03
jhollway/ to snlab-ch/ organisationElaborated documentation for the remainder of the datasets
graph_degree() where data was hard-coded inClosed #18 by adding blockmodel_concor() for employing the CONCOR algorithm to blockmodel both one-mode and two-mode networks
print.blockmodel() method in the sna package that also prints blockmodel results for two-mode networks consistently2021-02-06
as_matrix() function to coerce objects into an adjacency or incidence matrix classas_igraph() function to coerce objects into an igraph graph classas_tidygraph() function to coerce objects into an tidygraph tbl_graph classis_twomode() function to check whether network is two-mode on all object typesRenamed several datasets and elaborated their documentation
mpn_mexicanpower was renamed to mpn_elite_mex
mpn_powerelite was renamed to mpn_elite_usa_advice
mpn_opensecrets was renamed to mpn_elite_usa_money
Reconstructed several creation functions to take universal (one-mode/two-mode) input: specifying n = 5 creates a one-mode network, while specifying n = c(5, 5) creates a two-mode network
create_empty()
create_complete()
create_ring() to create rings of varying breadthcreate_components() (renamed from create_silos()) to create networks with varying numbers of componentssample_affiliation() for random two-mode networkscreate_match() and create_nest()
Renamed centrality_ functions with node_ prefix and ensured they all also wrapped one-mode measures
centrality_degree() renamed to node_degree()
centrality_closeness() renamed to node_closeness()
centrality_betweenness() renamed to node_betweenness()
node_eigenvector()
Re-added node_constraint() for calculating Burt’s constraint measure for one- and two-mode networks
Re-added node_smallworld() for calculating Watts-Strogatz measure of small-worldness for two-mode networks
Closed #32 by re-adding centralization functions for one- and two-mode networks
graph_degree() for degree centralizationgraph_closeness() for closeness centralizationgraph_betweenness() for betweenness centralizationRe-added graph_clustering() for calculating (see Knoke et al 2021):
plot.igraph() with sensible defaults for two-mode networks2021-01-08
project_rows() and project_cols() to make it easier to project two-mode networks in different formats (matrix, igraph, tidygraph) into projected versions in the same formatas_incidence_matrix()