Package 'netCoin'

Description

netCoin combines the data analysis capabilities of R with the interactive visualization libraries of JavaScript to create networks of coincidences, co-occurrences, correlations and regressions that can be explored directly in a browser, embedded in HTML files or integrated into Shiny applications. It also generates networked HTML galleries and other plots of coincidences.

Details

Coincidence analysis detects what events, characters, objects, attributes, or characteristics tend to occur together within certain limits.

These given limits are call scenarios ($S$) and are considered to be the units of analysis, and as such they have to be placed in the rows of a matrix or data.frame.

In each $i$ scenario, a series of $J$ events $X_j$, which are to be represented as dichotomous variables $X_{j}$ in columns, may occur (1) or may not occur (0). Scenarios and events constitute an incidence matrix (I).

Incidence matrix

From this incidences matrix, a coincidence (C) matrix can be obtained with the function coin. In this matrix the main diagonal represents frequencies of $X_j$, while the others elements are number of coincidences between two events.

Coincidence matrix

Once there is a coin object, a similarity matrix can be obtained. Similarity matrices available in netCoin are:

• Matching (m), Rogers & Tanimoto (t) Gower (g) Sneath (s) and Anderberg (and).

• Jaccard (j), dice (d), antiDice (a), Ochiai (o) and Kulczynski (k).

• Hamann (ham), Yule (y), Pearson (p), odds ratio (od) and Rusell (r).

Other measures that can be obtained from coin are:

• Relative frequencies (x), conditional frequencies (i) coincidence degree (cc) and probable degree of coincidence (cp).

• Haberman (h) and Z value of Haberman (z)

To obtain similarity and other measures matrices, the function sim elaborates a list of them.

Similarity matrix

edgeList makes a collecion of edges composed by a list of similarity measures whenever a criterium (generally p(Z)<.50) is met.

Edge list

In order to make a graph, two data frames are needed: a nodes data frames with names and other nodes attributes (see asNodes) and an edge data frame (see edgeList). For more information go to netCoin.

Author

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

References

Escobar, M. (2009): "Redes Semanticas en Textos Periodisticos: Propuestas Tecnicas para su Representacion", en Empiria, 17, 13-39.

Escobar, M.(2015): "Studying Coincidences with Network Analysis and Other Multivariate Tools", in The Stata Journal, 15(4), 1118-1156.

Escobar, M. and J. Gomez Isla (2015): "The Expression of Identity through the Image: The Photographic Archives of Miguel de Unamuno and Joaquin Turina", en Revista Espanola de Investigaciones Sociologicas, 152, 23-46.

Escobar, M. and L. Martínez-Uribe (2020): "Network Coincidence Analysis: The netCoin R Package", en Journal of Statistical Software, 93(11), 1-31.

Description

addAxes adds one or several planes of coordinates to a netCoin object from surScat, which the viewer then offers in a selector next to the graph.

Usage

addAxes(scatObj, axes, name=NULL, which=NULL, weight=NULL, axesLabels=NULL)

Arguments

Details

A scattergram drawn by surScat holds a single pair of coordinates, the one stated by its xaxis and yaxis arguments. addAxes adds further pairs to the same scattergram, so that the very same nodes, with their groups and their profiles, can be looked at from several projections without building a graph for each. The plane that was already drawn becomes the first entry of the selector, named after the axes it was drawn with.

Each pair of columns of axes makes up a plane: the first column is drawn horizontally and the second vertically. Planes are named after their two columns, as in "PC1-PC2", unless name states otherwise. A 'looking4clusters' object names them after its reductions ("pca", "tsne", "cmds", "nmf", "umap", plus any added through its own addreduction function). The 'looking4clusters' package needs not be installed: the object is read through its list structure.

The coordinates must come from the same data, in the same order, as the one given to surScat: only their length is checked, so keeping the order is up to the caller. Coordinates given case by case are collapsed into patterns, by their (weighted) mean, when surScat was called with vPatterns. The weights are the ones surScat itself collapsed with, which it keeps, unless weight states otherwise: a plane added to a weighted scattergram is therefore comparable with the one already drawn without having to repeat the weights.

They must therefore hold one row per node, or one per case when the nodes are patterns. Anything else is an error, which is also what happens when surScat capped the nodes through its maxN argument: the node table then holds a random sample of the cases, or of the patterns, so coordinates computed on the whole data no longer line up with it and cannot be matched to it. Compute the coordinates on the node table of the object, which is that sample, or raise maxN so that every case is drawn.

The labels of the axes belong to the graph and not to each plane, so they keep describing the plane that surScat drew. Naming the added planes after their axes keeps the reader oriented.

Every clusterization the object already holds (see surScat and addClusters) gets, in $clusters, a further entry for the plane just added, named by joining the clusterization and the plane with a dot, as in "kmeans(3).petals", so that $clusters always holds one entry per clusterization-plane pair, covering the same planes as $layouts.

Value

A 'netCoin' object.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

Examples

## Not run: 
data(iris)
S <- surScat(iris, variables=names(iris)[1:4], type="pca", nclusters=3)

# a second plane from further principal components
P <- prcomp(iris[,1:4], scale.=TRUE)$x
S <- addAxes(S, P[,3:4])

# every reduction of a looking4clusters object
# S <- addAxes(S, l4cObject)
plot(S)

## End(Not run)

Description

addClusters adds one or several cluster columns to a netCoin object from surScat.

Usage

addClusters(scatObj, clusters, name=NULL, sort=TRUE,
  weight=NULL, sourceData=NULL, which=NULL, maxGroups=25, rest=NULL)

Arguments

Details

Each column is an ordered factor whose levels follow the group number. Their labels are prefixed with the name of the clusterization they come from, capitalized and without its trailing number of groups, so a column named "hclust(10)" is labelled "Hclust: 01" ... "Hclust: 10". Group numbers are padded with zeros to the width of the largest one, which keeps them in numeric order wherever they are sorted as text.

Each clusterization is named after where it comes from and how many groups it holds, as in "kmeans(4)" for a method of a 'looking4clusters' object. The solutions of a 'tidyLPA' object are named "LPA(4)", and labelled "LPA: 1" ... "LPA: 4", since a latent profile analysis estimated for several numbers of profiles is one analysis. A 'tidyLPA' object estimated for several model specifications at once is no longer a single analysis, so its solutions carry the number tidyLPA gives each specification, as in "LPA1(4)" and "LPA6(4)".

Several clusterizations can be added in one call, either as the columns of a data.frame or as a 'looking4clusters' object. They must all come from the same data, in the same order, as the one given to surScat: only their length is checked, so keeping the order is up to the caller.

A clusterization must hold one assignment per node, or one per case when surScat collapsed the cases into patterns through vPatterns, in which case the cases of each pattern are reduced to their most frequent group. Anything else is an error, which is also what happens when surScat capped the nodes through its maxN argument: the node table then holds a random sample of the cases, or of the patterns, so a clusterization computed on the whole data no longer lines up with it and cannot be matched to it. Compute the clusters on the node table of the object, which is that sample, or raise maxN so that every case is drawn.

A 'looking4clusters' object usually holds many clusterizations at once: those added through its own addcluster function, plus one solution per number of groups for each method it has run. Leaving which empty brings all of them in, whereas which picks a single one. The 'looking4clusters' package needs not be installed: the object is read through its list structure.

Each clusterization added gets, in $clusters, one entry per plane the object holds (see surScat and addAxes), named by joining the clusterization and the plane with a dot, as in scatObj$clusters[["kmeans(3).petals"]]; each entry is a matrix with one row per group and its centroid coordinates. Where surScat collapsed the cases into patterns, a pattern spanning several groups keeps in its column the compound value (as in "Group: 1|Group: 2"), but there is still one centroid per group: the pattern takes part in the centroid of each group it spans with the (weighted) cases it holds of that group. These counts, one matrix per clusterization with a row per node and a column per group, are kept in the attribute "clusterShares" of the object, since the group of every case is no longer known once the cases are collapsed. $clusterMeans is updated as well, with the mean of every group of every clusterization on every numeric variable of the node table (see surScat). Printing $clusters, on its own or as part of the object, only states how many clusterizations and planes it covers.

Value

A 'netCoin' object.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

Description

addCommunities adds one cluster column per community detection method to a netCoin object from surScat, taken from the network of correlations between its cases.

Usage

addCommunities(scatObj, data, variables=names(data), community="lo",
  scale=TRUE, method=c("pearson", "kendall", "spearman"),
  criteria="value", minL=0.95, maxL=Inf, pairwise=FALSE,
  minSize=0.01, maxCommunities=NULL, name=NULL, sort=TRUE,
  weight=NULL, maxGroups=25, maxNodes=NULL, seed=2020)

Arguments

Details

The cases are correlated with one another over the variables, which is the transposition of the usual reading of a data.frame: each case becomes a column and each variable a row, so that a link joins two cases whose profiles run parallel. The communities of that network are a grouping of cases, of the same shape as any other clusterization, and reach the node table through addClusters, with the naming and the labelling described there. A louvain of five groups is a column named "louvain(5)", labelled "Louvain: 1" ... "Louvain: 5".

A correlation between two cases centres each of them on its own mean across the variables, so what it compares is the shape of their profiles. Standardizing the variables first puts them all on one footing; without it, the variable of widest range writes the profile on its own, which is why scale is on by default. Each correlation rests on as many points as there are variables, so comparing cases over very few of them says very little, and warns.

A network of cases leaves isolated ones behind, those reaching no other case above minL, and igraph returns each of them as a community of its own. minSize gathers those and any other small community into one residual group, which is pinned after the real ones and labelled with a word instead of a number, so that it reads apart from them in a legend. Without it a scattergram of three readable groups reports dozens and runs into maxGroups.

minSize sets a rule and lets the count of groups fall where it may; maxCommunities sets the count and lets the rule fall where it may, keeping the largest communities apart and gathering the rest, whatever their size. They work together, into one residual group, and a community falling to either of them ends up in it. Note that maxGroups is neither of these two: it gathers nothing, it merely refuses a clusterization of too many groups, so a maxCommunities at or below it means the column is always added.

Three of the methods draw on the random number generator, and so answer differently on the same network from one run to the next: louvain and label propagation sweep the nodes in a random order, and spinglass anneals from a random state. The other five (edge betweenness, fast greedy, leading eigenvector, optimal and walktrap) return the same communities every time. Left to run unseeded, the unstable three move their communities between runs, and with them the sizes minSize is read against, so that the same minSize gathers a different set of communities each time. seed therefore defaults to 2020, as in surScat, and is set afresh before each method, so that a method answers the same whichever others were asked for alongside it in community.

The whole network is built before the links are filtered, so its cost grows with the square of the number of cases: some thousands of them are a matter of seconds and a few gigabytes, and well beyond that it stops being feasible. The methods themselves do not all bear the same size either, and each one carries the ceiling it can meet: cluster_optimal is exponential and refuses above a hundred nodes, whereas cluster_louvain answers on thousands.

When surScat collapsed the cases into patterns through vPatterns, the communities are found case by case and then reduced to the most frequent one of each pattern, as any other clusterization is. When surScat capped the nodes through maxN, the network is built on the cases that were drawn, which the object records, so that the communities line up with the nodes that are actually plotted. The two together cannot be resolved: a sample of patterns cannot be traced back to the cases behind it, and that is an error.

Value

A 'netCoin' object.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

See Also

addClusters, surScat, netCorr.

Examples

## Not run: 
data <- data.frame(a=rnorm(300), b=rnorm(300), c=rnorm(300), d=rnorm(300))
S <- surScat(data, maxN=Inf)
S <- addCommunities(S, data, community=c("louvain", "walktrap"))
plot(S)

## End(Not run)

Description

addDescription adds a description to a 'netCoin' object.

Usage

addDescription(x, description)

Arguments

Value

A 'netCoin' object.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

Examples

data(finches)
data(Galapagos)

Net<-allNet(Galapagos,nodes=finches, criteria="hyp", maxL=.05,
        lwidth ="Haberman",lweight="Haberman",
        size="frequency", color="species", layout="mds",
        main="Species coincidences in Galapagos Islands",
        note="Data source: Sanderson (2000)")

img <- system.file("extdata", "p.Crassirostris.png",
        package="netCoin")
Net <- addDescription(Net,"Species coincidences in Galapagos Islands")

## Not run: 
multi <- multigraphCreate(Network=Net)
multiPages(multi,"Graph description example",show=TRUE)

## End(Not run)

Description

addImage adds an image to a 'netCoin' object.

Usage

addImage(x, img)

Arguments

Value

A 'netCoin' object.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

Examples

data(finches)
data(Galapagos)

Net<-allNet(Galapagos,nodes=finches, criteria="hyp", maxL=.05,
        lwidth ="Haberman",lweight="Haberman",
        size="frequency", color="species", layout="mds",
        main="Species coincidences in Galapagos Islands",
        note="Data source: Sanderson (2000)")

img <- system.file("extdata", "p.Crassirostris.png",
        package="netCoin")
Net <- addImage(Net,img)

## Not run: 
multi <- multigraphCreate(Network=Net)
multiPages(multi,"Graph image example",show=TRUE)

## End(Not run)

Description

addNetCoin changes netCoin object attributes.

Usage

addNetCoin(x, ...)

Arguments

Details

The object is rebuilt out of its own options, with the ones stated here replacing them, so anything not among the arguments of netCoin would be lost in the rebuilding and is carried over instead: the planes addAxes added, and what a scattergram records about how its nodes stand for the cases, which addClusters, addCommunities and replaceClusters read afterwards. Stating layout replaces the planes, as asked; replacing nodes with a table of another length carries none of it over, since none of it would describe those nodes.

Nodes drawn with no label at all, as surScat draws them, keep it that way. A label is restored by naming its column, as in label="name".

Value

This function returns a netCoin object. If the 'dir' attribute is specified, the function creates a folder in the computer with an HTML document named index.html which contains the produced graph. This file can be directly opened with your browser and sent to a web server to work properly.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

References

Escobar, M. and Martinez-Uribe, L. (2020) Network Coincidence Analysis: The netCoin R Package. Journal of Statistical Software, 93, 1-32. doi:10.18637/jss.v093.i11.

Examples

# A character column (with separator)
frame <- data.frame(A = c("Man; Women", "Women; Women",
                        "Man; Man", "Undet.; Women; Man"))
data <- dichotomize(frame, "A", sep = "; ")[2:4]
C <- coin(data) # coincidence matrix
N <- asNodes(C) # node data frame
E <- edgeList(C) # edge data frame
net <- netCoin(N, E) # netCoin object
net <- addNetCoin(net, color="frequency") # netCoin object with color applied to nodes

Description

addTutorial adds a tutorial for a gallery.

Usage

addTutorial(x, image = NULL, description = NULL, help = FALSE)

Arguments

Value

Object of class gallery_rd3.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

Examples

data("finches")
finches$species <- system.file("extdata", finches$species,
        package="netCoin") # copy path to the species field
g <- gallery(finches, image="species", main="Species in Galapagos Islands",
        note="Data source: Sanderson (2000)")

g <- addTutorial(g, 
  description="Here you can see different finches species in Galapagos islands.")

## Not run: 
plot(g)

## End(Not run)

Description

allNet produces a network object of coincidences from a data frame or a matrix with dichotomous values.

Usage

allNet(incidences, weight = NULL, subsample = FALSE, pairwise = FALSE,
       minimum=1, maximum = nrow(incidences), 
       sort = FALSE, decreasing = TRUE,
       frequency = FALSE, percentages = TRUE, 
       procedures = "Haberman", criteria = "Z", Bonferroni = FALSE,
       support = -Inf, minL = -Inf, maxL = Inf,
       directed = FALSE, diagonal = FALSE, 
       sortL = NULL, decreasingL = TRUE,
       igraph = FALSE, dir=NULL, ...)

Arguments

Details

Possible measures in procedures are

• Frequencies (f), Relative frequencies (x), Conditional frequencies (i), Coincidence degree (cc), Probable degree (cp),

• Expected (e), Confidence interval (con)

• Matching (m), Rogers & Tanimoto (t), Gower (g), Sneath (s), Anderberg (and),

• Jaccard (j), Dice (d), antiDice (a), Ochiai (o), Kulczynski (k),

• Hamann (ham), Yule (y), Pearson (p), odds ratio (od), Rusell (r),

• Haberman (h), Z value of Haberman (z),

• Hypergeometric p greater value (hyp).

• Convert a matrix into an edge list (shape).

Value

This function creates a netCoin object (or igraph) and, if stated, a folder in the computer with an HTML document named index.html which contains the produced graph. This file can be directly opened with your browser and sent to a web server to work properly.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

References

Escobar, M. and Martinez-Uribe, L. (2020) Network Coincidence Analysis: The netCoin R Package. Journal of Statistical Software, 93, 1-32. doi:10.18637/jss.v093.i11.

Examples

# A character column (with separator)
frame <- data.frame(A = c("Man; Women", "Women; Women",
                        "Man; Man", "Undet.; Women; Man"))
data <- dichotomize(frame, "A", sep = "; ")[2:4]
allNet(data) # network object

Description

asGallery produces a gallery_rd3 object.

Usage

asGallery(net)

Arguments

Value

Object of class gallery_rd3.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

Examples

data("Galapagos")
data("finches")
finches$species <- system.file("extdata", finches$species,
        package="netCoin") # copy path to the species field
Net <- allNet(Galapagos, nodes=finches, criteria="hyp", maxL=.05,
        image="species",
        main="Species coincidences in Galapagos Islands",
        note="Data source: Sanderson (2000)")
gallery <- asGallery(Net)

Description

Nodes data frame from either an edge list or a coin object.

Usage

asNodes(C, frequency = TRUE, percentages = FALSE, language = c("en","es","ca"))

Arguments

Value

A data frame with nodes' names and their frequency and/or percentages if the input is a coin object

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

References

Escobar, M. and Martinez-Uribe, L. (2020) Network Coincidence Analysis: The netCoin R Package. Journal of Statistical Software, 93, 1-32. doi:10.18637/jss.v093.i11.

Examples

# From a random incidence matrix I(25X4)
I <- matrix(rbinom(100, 1, .5), nrow = 25, ncol = 4,
     dimnames = list(NULL, c("A", "B", "C", "D")))
C <- coin(I)
asNodes(C)

Description

barCoin produces a barCoin object.

Usage

barCoin(data, variables = colnames(data), commonlabel = NULL,
        dichotomies = c("_all","_none"), valueDicho = 1, weight = NULL,
        subsample = FALSE, sort = NULL, decreasing = TRUE, nodes = NULL,
        name = NULL, select = NULL, scalebar = FALSE, note = NULL, 
        label = NULL, text = NULL, color = NULL, defaultColor = "#1f77b4", 
        expected = FALSE, confidence = FALSE, level = .95, significance = FALSE, 
        minimum = 1 , maximum = nrow(data), percentages = FALSE, 
        criteria = c("Z","hyp"), Bonferroni = FALSE, 
        support = 1, minL = -Inf, maxL = 1,
        language = c("en","es","ca"), cex = 1.0, dir = NULL)

Arguments

Value

Object of class barCoin.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

Examples

# A character column (with separator)
frame <- data.frame(A = c("Man; Women", "Women; Women",
                        "Man; Man", "Undet.; Women; Man"))
data <- dichotomize(frame, "A", sep = "; ")[2:4]

barCoin(data,dichotomies="_all")
barCoin(data,dichotomies="_all",confidence=TRUE,percentages=TRUE)

Description

This function calculates the centrality measures of a network.

Usage

calCentr(graph,
  measures = c("degree","wdegree","closeness","betweenness","eigen"),
  order = "")

Arguments

Details

This function reproduces some of the most significant classic Social Network Theory's centrality measures. See Wasserman (1994), Freeman (1978), or Bonacich & Lloyd (2001) to know more.

a) Degree = Degree centrality is measured by the total amount of direct links with the other nodes.

b) Closeness = Closeness centrality is meant to measure one node to the others nodes' sum distances

c) Betweenness = Betweenness centrality measures one node undertaking "mediation" role in a network.

d) Eigen = Eigenvector centrality measures a node's importance while giving consideration to the importance of its neighbors.

By default, measures = "all", thus all the measures will be calculated. The function can be applied to an igraph or a netCoin object. In case the graph is undirected, it will show the degree, weighted degree, closeness, betweeness and eigen degree. Moreover, if it us directed, it will show the indegree, windegree and outdegree, both weighted and unweighted (See example.).

Value

This function creates a list containing two elements: 1) a data.frame (nodes) with all the centrality measures applied to the graph and 2) another data.frame (graph) with this measures applied to the whole network.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

Examples

# A character column (with separator)
frame <- data.frame(A = c("Man; Women", "Women; Women",
                        "Man; Man", "Undet.; Women; Man"))
data <- dichotomize(frame, "A", sep = "; ")[2:4]
C <- coin(data) # coincidence matrix
N <- asNodes(C) # node data frame
E <- edgeList(C) # edge data frame
G <- netCoin(N, E, showArrows=TRUE) # netCoin object
calCentr(G, "all")

Description

caring produce an interactive multi graph from caring arguments.

Usage

caring(filepath,arguments)

Arguments

Value

This function returns a mGraph object. The function creates a folder in your computer with an HTML document named index.html which contains the graph. This file can be directly opened with your browser.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

Examples

## Not run: 
filename <- "demo.sav"
arguments <- list(
  variables = c("Gender", "Income category in thousands",
    "Primary vehicle price category"),
  plot = "network"
)
graph <- caring(filename,arguments)
plot(graph)

## End(Not run)

Description

caring_create_graphs produce an interactive multi graph from caring arguments.

Usage

caring_create_graphs(data,arguments)

Arguments

Value

This function returns a mGraph object. The function creates a folder in your computer with an HTML document named index.html which contains the graph. This file can be directly opened with your browser.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

Examples

data <- data.frame(Gender=c(rep("Man",3),rep("Woman",3)),
                    Opinion=c("Yes","Yes","No","No","No","Yes"))
arguments <- list(
  variables = c("Gender", "Opinion"),
  dichotomies = "Opinion",
  valueDicho = "Yes",
  plot = c("surCoin","network")
)
graph <- caring_create_graphs(data,arguments)
## Not run: 
plot(graph)

## End(Not run)

Description

caring_read_file reads sav, dta, xlsx, tsv and csv files and prepares data for caring.

Usage

caring_read_file(filepath)

Arguments

Value

This function returns a data frame.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

Examples

## Not run: 
filename <- "demo.sav"
data <- caring_read_file(filename)

## End(Not run)

Description

clusterAgree measures the agreement between a model classification and the other classifications held by a netCoin object from surScat through ARI, NMI and VI, which match no categories and take therefore classifications with any number of groups.

Usage

clusterAgree(x, model, columns=NULL, weights=NULL, modal=TRUE, sort=NULL, digits=3)

Arguments

Details

Splitting finer never harms precision — in the limit, one group per case leaves it at 1 — so the measures of a confusion matrix cannot arbitrate between classifications with different numbers of groups, and clusterMatch refuses to compare them at all. These three indices are there for that: they read the two classifications as partitions and need no matching of categories.

ARI, the adjusted Rand index, walks through the pairs of cases, counting whether the two classifications put them together or keep them apart, and corrects for chance: 0 is what independent partitions yield, 1 an exact match. It may go slightly negative.

NMI, the normalised mutual information, tells how much one classification cuts down the uncertainty about the other, from 0 to 1. It is not corrected for chance, so it tends to rise with the number of categories and does not arbitrate well between solutions of different size.

VI, the variation of information, is a distance in nats: 0 is an exact match and it grows with the discrepancy. VI.n divides it by log(N) to keep it between 0 and 1. Note when sorting that in these two the small value is the good one, so sort="-VI" is usually what is wanted.

Where the nodes are patterns, weighting and modal groups work just as in clusterMatch. Weights are taken to be on a scale of cases, which is what the default weighting yields: the count of pairs behind ARI is extended continuously, and weights given as proportions would make it meaningless.

Value

A data.frame with one row per classification, sorted as sort states. It carries as attributes the contingency table of every classification ("tables"), the number of categories of the model ("kModel"), the unit of analysis ("unit"), the weights actually applied ("weights") and, where there were any, the patterns spanning several groups ("spanning").

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

See Also

clusterMatch, surScat, addClusters

Examples

## Not run: 
data(iris)
S <- surScat(iris, variables=names(iris)[1:4], type="pca", nclusters=2:4)

# every classification, whatever its number of groups
A <- clusterAgree(S, "Groups(3)", sort="ARI")
A
attr(A, "tables")[[1]]

# the most concordant on top, VI being a distance
clusterAgree(S, "Groups(3)", sort="-VI")

## End(Not run)

Description

clusterMatch compares a model classification with the other classifications held by a netCoin object from surScat, matching their categories one to one and reporting sensitivity, precision, F1 and specificity.

Usage

clusterMatch(x, model, columns=NULL, weights=NULL, align=TRUE, weighted=TRUE,
             modal=TRUE, sort=NULL, digits=3)

Arguments

Details

The groups of a classification carry no name of their own: group 1 of one method need not be group 1 of another, as the numbering is arbitrary. Before comparing, the categories of every classification are therefore realigned with those of the model through the optimal assignment (Hungarian algorithm), which maximizes the diagonal of the contingency table. That demands a one-to-one correspondence, and clusterMatch compares therefore only the classifications holding as many categories as the model: without a bijection these measures are not defined. clusterAgree is there for the rest.

The measures follow the definitions of caret::confusionMatrix (one against the rest, prediction down the rows and reference along the columns) but are computed here so as not to depend on caret. Every one of them is averaged over the categories in two ways: the plain mean treats them alike, and a small group badly classified sinks it; the one weighted by the size of each group measures what happens to a case drawn at random. Sensitivity is given no weighted mean, as it would coincide with Accuracy.

Where surScat has collapsed the cases into patterns, every node stands for several cases and counting it once would give the rare pattern the same say as the common one. The nodes are therefore weighed by the cases they stand for (the "caseToPattern" attribute, and "caseWeight" where surScat was given weights), so that the tables come out in cases.

Pattern nodes bring a limitation of the object itself: surScat keeps in every cluster column all the groups the pattern spans, as text ordered from the commonest down ("Group: 2|Group: 1"), and loses which group each case went to. The comparison is then made with the modal group of every pattern, and how many patterns span more than one is reported: where they are many, the figures are optimistic and it is worth comparing case by case, passing the model as a vector of cases and classifications that have not been collapsed.

Value

A data.frame with one row per classification, sorted as sort states. It carries as attributes the confusion matrix of every classification ("tables"), its measures category by category ("byClass"), the matching of categories ("mapping"), the unit of analysis ("unit"), the weights actually applied ("weights") and, where there were any, the patterns spanning several groups ("spanning").

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

See Also

clusterAgree, surScat, addClusters

Examples

## Not run: 
data(iris)
S <- surScat(iris, variables=names(iris)[1:4], type="pca", nclusters=2:4, clusters="hclust")

R <- clusterMatch(S, "Groups(3)")
R
attr(R, "byClass")[[1]]    # measures category by category
attr(R, "tables")[[1]]     # the realigned confusion matrix
attr(R, "mapping")[[1]]    # which category was matched with which

# sorted by F1, and by the simplest classifications first
clusterMatch(S, "Groups(3)", sort="F1")
clusterMatch(S, "Groups(3)", sort=c("-k", "F1"))

# with an outside vector rather than a node column
clusterMatch(S, iris$Species)

## End(Not run)

Description

cobCoin produces a netCoin object to graph a CoWeb graphic (Upton 2000).

Usage

cobCoin(data, variables=names(data), degree=0, significance=.05, ...)

Arguments

Value

This function creates a netCoin object (or igraph) and, if stated, a folder in the computer with an HTML document named index.html which contains the produced graph. This file can be directly opened with your browser and sent to a web server to work properly.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

References

Upton, Graham J.G. (2000) Cobweb diagrams for multiway contingency tables. The Statistician, 49(1), 79-85.

Examples

# A data frame with two variables Gender and Opinion
frame <- data.frame(Gender=c(rep("Man",3),rep("Woman",3)),
                    Opinion=c("Yes","Yes","No","No","No","Yes"))
cobCoin(frame, significance=.5, degree=45) # netCoin object

# A data frame with two variables (Gender and Hand) and nodes
input <- data.frame(
  Gender = c("Women", "Men", "Men", "Women", "Women","Men",
             "Men", "Men", "Women", "Women", "Men", "Women"),
  Hand   = c("Right", "Left","Right", "Right", "Right", "Right",
             "Left", "Right", "Right", "Left","Right", "Right"))
nodes <- data.frame(
  name  = c("Gender:Men","Gender:Women", "Hand:Left", "Hand:Right"),
  label = c("Women(50\u25)","Men(50\u25)",
            "Left hand(25\u25)", "Right hand(75\u25)"))
G <- cobCoin(input, nodes=nodes, label="label", degree=22.5,
             showArrows=TRUE, significance=1)

Description

Builds a side-by-side coefficient table from one or several objects returned by glmCoin. It reads the coeffs list stored in them (one data frame of coefficients per equation) and the equations data frame (n and pseudo-R2 of each model). Note that it does not read links, which holds average marginal effects rather than coefficients: that is what marginsTable tabulates. Each column block corresponds to one combination of object, equation and Target, so the same function covers hierarchical regressions (one response, nested sets of predictors), several responses in a single object, multinomial responses (one block per category) and any combination of these.

Usage

coefTable(G, labels = NULL, columns = c("coef", "sig", "se"),
          stfactors = c("nostandard", "x", "long"),
          blocks = c("entry", "variable", "none"), intercept = FALSE,
          sig = c(0.05, 0.01, 0.001), digits = 3, digits.se = digits,
          footer = c("n_obs", "r2", "dr2"), note = FALSE,
          file = NULL, format = NULL)

Arguments

Details

Rows are the union of the predictors of all the equations, in order of first appearance; an empty cell means that the predictor was not included in that equation, which is not the same as a zero coefficient.

Because coeffs and equations keep the same order, the columns are matched to their fit statistics by position, without comparing model strings.

Because R2 and its increments are only comparable when the equations are fitted on the same cases, a warning is issued whenever the models differ in their number of observations, and the fact is recorded in the note when note=TRUE. This typically happens when a predictor added in a later step has missing values.

Value

A character matrix of class coefTable, printed with grouped headers (one block of columns per model, response or category). The bottom rows show the statistics selected in footer. When exporting, the table is returned invisibly.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

See Also

glmCoin, marginsTable

Examples

data(birthwt, package="MASS")
birthwt$smoke <- factor(birthwt$smoke, labels=c("no","yes"))
birthwt$race  <- factor(birthwt$race,  labels=c("white","black","other"))

## Hierarchical regression: one response, nested sets of predictors.
## Blocks are deduced from the equation in which each predictor enters.
m1 <- glmCoin("bwt ~ age + lwt", birthwt)
m2 <- glmCoin("bwt ~ age + lwt + smoke", birthwt)
m3 <- glmCoin("bwt ~ age + lwt + smoke + race", birthwt)
coefTable(list(m1, m2, m3), note=TRUE)

## Standardized coefficients: quantitative terms only, then all of them
coefTable(m3, columns=c("coef","sta"))
coefTable(m3, columns=c("coef","sta"), stfactors="x")

## Confidence intervals instead of standard errors, with fewer decimals
coefTable(list(m1, m2, m3), columns=c("coef","sig","ci"), digits=2, digits.se=1)

## Logistic regression: odds ratios beside the coefficients
g <- glmCoin(c("bwt ~ age + lwt", "low ~ age + lwt, binomial"), birthwt)
coefTable(g, columns=c("coef","sig","exp"), footer=c("n_obs","r2"))

## Multinomial response: one block of columns per category
data(housing, package="MASS")
h <- glmCoin("Sat ~ Infl + Type + Cont, multinomial", housing, weights="Freq")
coefTable(h, blocks="variable", columns=c("coef","sig"))

Description

coexist produces interactive graphs representing coexistence. Two periods or lifes coexist if they share a given number of years.

Usage

coexist(periods, name="name", start="start", end="end", 
        fields=names(periods), plusstart=0, minusend=0, igraph=FALSE, ...)

Arguments

Details

Two periods coexists if they have at least one year in common. Periods can be trimmed at the begining or at the end.

Value

This function creates a netCoin object (or igraph) and, if plotted, a folder in the computer with an HTML document named index.html which contains the produced graph. This file can be directly opened with your browser and sent to a web server to work properly.

A netCoin object has three elements:

Note

Periods could be the life of people, in whose case start is their birth and end their death year.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

See Also

timeCoin and dyncohort

Examples

# With sociologists data.
data(sociologists)
# Produce an igraph object:
coexist(sociologists, start="birth", end="death", igraph=TRUE)
# Omit as coexistence the 15 firts years of life:
sociologists$picture <- system.file("extdata", sociologists$picture,
        package="netCoin") # copy path to the picture field
coexist(sociologists, start="birth", end="death", plusstart=15, image="picture", imageName="name")

Description

A coin object is a squared matrix of coincidences with its names and the frequency of events as attributes.

Usage

coin(incidences, minimum = 1, maximum = nrow(incidences),
     sort = FALSE, decreasing = TRUE, 
     total = FALSE, subsample = FALSE,
     weight = NULL, pairwise = FALSE)

Arguments

Details

Produce a matrix of coincidences from a matrix of incidences.

Value

An object of coin class

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

References

Escobar, M. and Martinez-Uribe, L. (2020) Network Coincidence Analysis: The netCoin R Package. Journal of Statistical Software, 93, 1-32. doi:10.18637/jss.v093.i11.

Examples

## Random incidence matrix: 25 scenarios, 4 events.
I <- matrix(rbinom(100, 1, .5), nrow = 25, ncol = 4,
            dimnames = list(NULL, c("A", "B", "C", "D")))
coin(I, sort = TRUE)

## Hair by Eye by Sex table from M. Friendly (2000)
data(HairEyeColor)
H<-as.data.frame(HairEyeColor)
W<-H$Freq
I<-dichotomize(H,c("Hair","Eye","Sex"),add=FALSE)
coin(I,w=W)

Description

A coocurrence object consists of a matrix with the number of ocurrences in its main diagonal and the number of coocurrences outside this diagonal. Besides, this object has two attributes: 1) n is the total of the sum of the ocurrences in each row.2) m is the sum of the maximum number of ocurrences in each row.

Usage

coocur(ocurrences, minimum = 1, maximum = Inf,
       sort = FALSE, decreasing = TRUE)

Arguments

Details

Produce a matrix of coocurrences from a matrix of occurences.

Value

An object of cooc class with a coocurrence matrix. It has two attributes:

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

Examples

## Tossing two coins five times.
D<-data.frame(Head=c(2,1,1,0,2),Tail=c(0,1,1,2,0))
coocur(D)

Description

netCorr produces a network object of dynamic correlations. Its input has to be at least one set of quantitative variables.

Usage

d_netCorr(variables, nodes = NULL, weight=NULL, 
        pairwise=FALSE, minimum=-Inf, maximum=Inf,
        frequency=FALSE, means=TRUE, 
        method=c("pearson", "kendall", "spearman"), criteria="value", Bonferroni=FALSE,
        minL=0, maxL=Inf,
        sortL=NULL, decreasingL=TRUE,
        factorial=c("null", "pc", "nf", "vf", "of"),
        components=TRUE, backcomponents=FALSE,
        sequence=seq(.20, 1, .01), textFilter=c(1, .99), speed=50,
        dir=NULL, ...)

Arguments

Value

The function creates a list object to create a multigraph and eventually a folder in the computer with an HTML document named index.html which contains the produced dynamic graphs. This file can be directly opened with your browser and sent to a web server to work properly.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

Examples

# A character column (with separator)
data(iris)
d_netCorr(iris[,1:4],ltext="value", sequence=seq(.80, 1, .05), zoom=3, textFilter=.85,
 main="Correlations between measurements of Iris Species", components=FALSE,
 note="Anderson, Edgar (1935) y Fisher, R. A. (1936)") # network object

Description

Data frame with events as result.

Usage

data("dice")

Format

A data frame with 100 observations (scenarios) on the following 11 variables (events):

dice

: a numeric vector, representing dice results

1

: a dichotomous vector of the elemental event "1"

2

: a dichotomous vector of the elemental event "2"

3

: a dichotomous vector of the elemental event "3"

4

: a dichotomous vector of the elemental event "4"

5

: a dichotomous vector of the elemental event "5"

6

: a dichotomous vector of the elemental event "6"

odd

: a dichotomous vector of odd events

even

: a dichotomous vector of even events

small

: a dichotomous vector of small number events

large

: a dichotomous vector of large number events

Source

Random extraction via sample(1:6,100,replace=TRUE)

References

See events.

Examples

data(dice)
head(dice,10)

Description

Converts factor or character columns of a data frame into a set of dichotomous (dummy/indicator) columns. Their names will correspond to the unique labels or texts of every category discovered across the selected variables. If multiple columns are provided, categories are unified per row.

Usage

dichotomize (data, variables, sep = "", regex = FALSE, 
               text = c("none", "lower", "upper"),
               min = 1, length = 0, sort = TRUE,
               values = NULL, stopwords = NULL, nas = "None",
               sparse = FALSE, add = FALSE)

Arguments

Value

A data frame composed of the original variables plus the added dichotomous columns, or a sparse matrix if sparse = TRUE.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca, and Luis Martinez Uribe, Fundacion Juan March.

References

Escobar, M. and Martinez-Uribe, L. (2020) Network Coincidence Analysis: The netCoin R Package. Journal of Statistical Software, 93, 1-32. doi:10.18637/jss.v093.i11.

Examples

# A character column
  frame1 <- data.frame(A = c("Man", "Women", "Man", "Undet."))
  dichotomize(frame1, "A", sep = "; ")
  
  # A character column (with separator)
  frame2 <- data.frame(A = c("Man; Women", "Women; Women",
                             "Man; Man", "Undet.; Women; Man"))
  dichotomize(frame2, "A", sep = "; ")
  
  # A character column and another factor column (same separator)
  frame3 <- data.frame(A = c("Man; Women", "Women; Women",
                             
                             "Man; Man", "Undet.; Women; Man"),
                       C = factor(c(1:4), labels = c("Paris", "New York",
                                                     "London; New York", "<NA>")))
  dichotomize(frame3, c("A", "C"), sep = "; ")
  
  # A set of simple character or factor (same levels) variables.
  # In this case, you must use "C" separator.
  frame4 <- data.frame(A = c("Man", "Women","Man", "Undet",NA),
                       B = c("Women","Women","Man","Women",NA),
                       C = c(NA,NA,NA,"Man",NA))
  dichotomize(frame4, c("A","B","C"), sep="C")
  
  # Text mining example: Splitting by whitespace, removing punctuation/numbers, 
  # transforming to lowercase and filtering stopwords
  frame5 <- data.frame(text_col = c("I love apples and 10 oranges!", 
                                    "Apples are great.", 
                                    "I hate bananas."))
  dichotomize(frame5, "text_col", sep = "\\s+", regex = TRUE, 
              text = "lower", stopwords = c("i", "and", "are"))

Description

Convert a similarity matrix into a distance matrix.

Usage

distant(s, t = FALSE)

Arguments

Details

For better resultas, use the parameter distance in sim function.

Value

A distance matrix.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

Examples

# From a random incidence matrix I(25X4)
I <- matrix(rbinom(100, 1, .5), nrow = 25, ncol = 4,
     dimnames = list(NULL, c("A", "B", "C", "D")))
J <- sim(I, "Jaccard")
distant(J, t = TRUE)
#Same results
sim(I, "Jaccard", distance = TRUE)

Description

dyncohort produces interactive graphs representing dynamic cohorts. Two periods or lifes belongs to the same cohort if there are a difference of years in their start less or equal to a given number. In case of people's life, 15 or 25 are appropiate quantities to set. If year is equal to 0, a cohort is defined a those periods or lifes that begin at the same year.

Usage

dyncohort(periods, name="name", start="start", fields=names(periods), 
          years=0, igraph=FALSE, ...)

Arguments

Value

This function creates a timeCoin object (or igraph) and, if plotted, a folder in the computer with an HTML document named index.html which contains the produced graph. This file can be directly opened with your browser and sent to a web server to work properly.

A netCoin object has three elements:

Note

Periods could be the life of people, in whose case start is their birth year.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

See Also

timeCoin and coexist

Examples

# With sociologists data.
data(sociologists)
dyncohort(sociologists, start="birth")
# Condider 15 year dynamic cohort:
sociologists$picture <- system.file("extdata", sociologists$picture,
        package="netCoin") # copy path to the picture field
dyncohort(sociologists, start="birth", year=15, image="picture", imageName="name")

Description

Convert a coincidence/similarity/distance matrix into an edge list form.

Usage

edgeList(data, procedures="Haberman", 
         criteria="Z", level=.95, Bonferroni=FALSE, 
         min=-Inf, max=Inf, support=-Inf, directed=FALSE, 
         diagonal=FALSE, sort=NULL, decreasing=TRUE, pairwise=FALSE)

Arguments

Details

Possible measures in procedures are

• Frequencies (f), Relative frequencies (x), Conditional frequencies (i), Coincidence degree (cc), Probable degree (cp),

• Expected (e), Confidence interval (con)

• Matching (m), Rogers & Tanimoto (t), Gower (g), Sneath (s), Anderberg (and),

• Jaccard (j), Dice (d), antiDice (a), Ochiai (o), Kulczynski (k),

• Hamann (ham), Yule (y), Pearson (p), odds ratio (od), Rusell (r),

• Haberman (h), Z value of Haberman (z),

• Hypergeometric p greater value (hyp).

• Convert a matrix into an edge list (shape).

Value

A data frame in which the two first columns are source and target. The rest of the columns are the different statistics explicited in funcs parameter.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

References

Escobar, M. and Martinez-Uribe, L. (2020) Network Coincidence Analysis: The netCoin R Package. Journal of Statistical Software, 93, 1-32. doi:10.18637/jss.v093.i11.

Examples

# From a random incidence matrix I(25X4)
I<-matrix(rbinom(100,1,.5),nrow=25,ncol=4,
dimnames=list(NULL,c("A","B","C","D")))
C<-coin(I)
edgeList(C)

Description

A sample size of 1,000 respondents from the European Social Survey, Round-8.

Usage

data("ess")

Format

A data frame with 1000 cases (respondents) and 5 variables:

Gender

Gender (factor vector): Female, Male.

Age

Age (recoded factor vector): 15-29, 30-30, 40-49, 50-59, 60-69, 70 and +.

Social participation

Social participation (factor vector): No, Yes.

Political participation

Political participation (factor vector): No, Yes.

cweight

cweight (numeric vector): Cases weight.

References

ESS Round 8: European Social Survey Round 8 Data (2016). Data file edition 2.1. NSD - Norwegian Centre for Research Data, Norway - Data Archive and distributor of ESS data for ESS ERIC. doi:10.21338/NSD-ESS8-2016.

Examples

data("ess")
head(ess,10)

Description

Data frame with the attributes of the events of dice.

Usage

data("events")

Format

A data frame with 10 observations on the following 4 variables:

name

: a factor vector with 10 levels

label

: a factor vector with 10 levels

frequency

: a numeric vector

type

: a factor vector with 2 levels

Source

data(dice); coin.dice<-coin(dice); asNodes(coin.dice)

References

See dice.

Examples

data(events)
events

Description

exhibit produces an interactive image gallery (alternative display).

Usage

exhibit(nodes, name = NULL, label = NULL, subtitle = NULL, order = NULL,
    decreasing = FALSE, ntext = NULL,
    mainframeHeight = NULL, mainframeWidth = NULL, mainframeImage = 0,
    image = NULL, imageCopy = NULL, noFilterCols = NULL,
    zoom = NULL, scale = NULL, main = NULL, note = NULL,
    logo = TRUE, pages = c("bottom","dynamic","top"), mainposition = c("dynamic","fixed"),
    frequencies = FALSE, tutorial = FALSE, tableButton = FALSE, export = FALSE,
    search = TRUE, colorScheme = 0,  language = c("en", "es", "ca"), dir = NULL)

Arguments

Value

Object of class gallery_rd3.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

Examples

data("finches")
finches$species <- system.file("extdata", finches$species,
        package="netCoin") # copy path to the species field
exhibit(finches, image="species")

Description

Converts a coin object to a links data frame with coincidences and expected values.

Usage

expectedList(data, names = NULL, min = 1, confidence=FALSE)

Arguments

Value

A links data frame with coincidences and expected values.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

Examples

# A character column (with separator)
frame <- data.frame(A = c("Man; Women", "Women; Women",
                        "Man; Man", "Undet.; Women; Man"))
data <- dichotomize(frame, "A", sep = "; ")[2:4]
C <- coin(data) # coincidence matrix
expectedList(C) # edge data frame

Description

Data frame with the characteristics of powerful families of Renaissance Italy.

Usage

data("families")

Format

A data frame with 16 families (rows) and 6 characteristics.

name

Family's name

f.Marriages

number of marriage links

f.Business

number of business links

wealth

wealth's index

priorates

number of priorates on control

seat

At least priorate

Source

PADGETT, J. F. Y C. K. ANSELL (1993): "Robust Action and the Rise of the Medici, 1400-1434", in American Journal of Sociology, 98, 1259-1319. (http://www.jstor.org/stable/2781822)

Examples

data("families")
head(families)

Description

Data frame with events as result.

Usage

data("finches")

Format

A data frame with 13 observations (pinches) and 4 variables (name and characteristics):

name

: Genus and species of the finche

frequency

: number of islands where the finche can be found

type

: Genus of the finche

species

: name of the file containing the picture of the finche

References

Sanderson, James (2000). Testing Ecological Patterns: A Well-known Algorithm from Computer Science Aids the Evaluation of Species Distributions. American Scientist, 88, pp. 332-339.

Examples

data(finches)
head(finches,10)

Description

fromIgraph produce an interactive network from an igraph object.

Usage

fromIgraph(G, ...)

Arguments

Value

This function returns a netCoin object. If the 'dir' attribute is specified, the function creates a folder in the computer with an HTML document named index.html which contains the produced graph. This file can be directly opened with your browser and sent to a web server to work properly.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

References

Escobar, M. and Martinez-Uribe, L. (2020) Network Coincidence Analysis: The netCoin R Package. Journal of Statistical Software, 93, 1-32. doi:10.18637/jss.v093.i11.

Examples

g <- igraph::make_ring(10)
fromIgraph(g)

Description

Data frame with absence(0) presence(1) of finches in the Galagos Islands.

Usage

data("Galapagos")

Format

A data frame with 17 localizations (islands) and 13 variables (Genus and species of the finches):

Geospiza magnirostris

Geospiza fortis

Geospiza fuliginosa

Geospiza difficilis

Geospiza scandens

Geospiza conirostris

Camarhynchus psitticula

Camarhynchus pauper

Camarhynchus parvulus

Platyspiza crassirostris

Cactospiza pallida

Cactospiza heliobates

Certhidea olivacea

References

Sanderson, James (2000). Testing Ecological Patterns: A Well-known Algorithm from Computer Science Aids the Evaluation of Species Distributions. American Scientist, 88, pp. 332-339.

Examples

data(Galapagos)
head(Galapagos,10)

Description

gallery produces an interactive image gallery.

Usage

gallery(nodes, name = NULL, label = NULL, color = NULL,
    border = NULL, ntext = NULL, info = NULL, infoFrame = c("right","left"),
    image = NULL, imageCopy = NULL, noFilterCols = NULL, zoom = NULL,
    scale = 1, itemsPerRow = NULL, main = NULL, note = NULL,
    showLegend = TRUE, frequencies = FALSE, labelTooltip = TRUE,
    cexTooltip = 1, help = NULL, helpOn = FALSE, tutorial = FALSE,
    description = NULL, descriptionWidth = NULL, roundedItems = FALSE,
    ntextctrl = FALSE, controls = 1:5, cex = 1, defaultColor = "#1f77b4",
    language = c("en", "es", "ca"), dir = NULL)

Arguments

Value

Object of class gallery_rd3.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

Examples

data("finches")
finches$species <- system.file("extdata", finches$species,
        package="netCoin") # copy path to the species field
gallery(finches, image="species", main="Species in Galapagos Islands",
        note="Data source: Sanderson (2000)")

Description

gallery3 produces an interactive image gallery (alternative display).

Usage

gallery3(nodes, name = NULL, label = NULL, subtitle = NULL, order = NULL,
    decreasing = FALSE, ntext = NULL,
    mainframeHeight = NULL, mainframeWidth = NULL, mainframeImage = 0,
    image = NULL, imageCopy = NULL, imageRatio = NULL, noFilterCols = NULL,
    zoom = NULL, scale = NULL, main = NULL, note = NULL,
    tableButton = FALSE, export = FALSE,
    search = TRUE, cex = 1, language = c("en", "es", "ca"), dir = NULL)

Arguments

Value

Object of class gallery_rd3.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

Examples

data("finches")
finches$species <- system.file("extdata", finches$species,
        package="rD3plot") # copy path to the species field
gallery <- gallery3(finches, image="species")
## Not run: 
plot(gallery)

## End(Not run)

Description

Create an html panel for nodes from different items (for galleries).

Usage

get_panel_template(data, title=NULL, description=NULL, img=NULL,
                   text=NULL, color="auto", cex=1, mode = 1)

Arguments

Value

a character vector of html formatted panel.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

Examples

## Not run: 
data("sociologists")
sociologists$picture <- system.file("extdata", sociologists$picture,
        package="netCoin")
sociologists$html <- get_panel_template(sociologists, title="name",
  description="birth_country",img="picture", text="school")
plot(gallery(sociologists, info="html"))

## End(Not run)

Description

Create a drop-down vignette for nodes from different items (for galleries).

Usage

get_template(
  data,
  title = NULL,
  title2 = NULL,
  text = NULL,
  img = NULL,
  wiki = NULL,
  width = 300,
  color = "auto",
  cex = 1,
  roundedImg = FALSE,
  mode = 1
)

Arguments

Value

a character vector of html formatted vignettes.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

Examples

## Not run: 
data("sociologists")
sociologists$picture <- system.file("extdata", sociologists$picture,
        package="netCoin")
sociologists$html <- get_template(sociologists, title="name",
  title2="birth_country", text="school",img="picture", mode=2)
plot(gallery(sociologists, ntext="html"))

## End(Not run)

Description

Create a drop-down vignette for nodes from different items (for galleries).

Usage

get_template2(data, title=NULL, title2=NULL, text=NULL, wiki=NULL)

Arguments

Value

a character vector of html formatted vignettes.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

Examples

## Not run: 
data("sociologists")
sociologists$html <- get_template2(sociologists, title="name",
  title2="birth_country", text="school")
plot(exhibit(sociologists, ntext="html"))

## End(Not run)

Description

produces a netCoin object from a set of glm regressions.

Usage

glmCoin(formulas, data, weights=NULL, pmax=.05,
        robust=FALSE, showArrows=TRUE,
        color="variable", lwidth="s.value", lcolor="estimate",
        language=c("en","es","ca"),
        igraph=FALSE, ...)

Arguments

Value

This function creates a netCoin object (or igraph) and, if stated, a folder in the computer with an HTML document named index.html which contains the produced graph. This file can be directly opened with your browser and sent to a web server to work properly.

The links of the graph carry the average marginal effects as GW contrasts (estimate, std.error, p.value: effects relative to the weighted overall mean) and also relative to the first category of each qualitative variable (estimate.base, std.error.base, p.value.base; for the base category itself the estimate is 0 and its standard error and p value are NA). Both sets can be displayed with marginsTable.

The netCoin object also includes a data frame named equations with one row per declared equation and the following goodness-of-fit measures: model (formula), family (declared family), n (number of observations; the sum of weights in weighted models), n_obs (number of observed cases, regardless of weights), r2 (deviance-based R2: 1-deviance/null deviance, which equals the classical R2 for gaussian models), r2.mcfadden (McFadden's pseudo-R2, NA for quasi families), r2.nagelkerke (Nagelkerke's pseudo-R2), logLik, aic, bic (NA for quasi families), deviance, null.deviance and df.residual. When igraph=TRUE, this data frame is attached as the graph attribute equations.

It also includes a list named coeffs with one data frame of regression coefficients per equation, in the same order as the rows of equations. Each data frame has one row per coefficient and response category, with columns Target (the response, or response:category in binomial and multinomial models), term (the name given by the fitted model), Source (the same term written as a node name, so that it can be matched against nodes and links), factor (whether the term involves a qualitative variable), estimate, std.error, statistic, p.value, conf.low, conf.high and sd.x (the weighted standard deviation of the term, NA for the constant). The attributes family, link, sd.y (the weighted standard deviation of the response, NA unless it is on its own scale) and latent.sd (the standard deviation of the underlying latent variable, per Target, NA unless the link is logit or probit) record the scales needed to standardize or exponentiate the estimates. These are the coefficients that coefTable displays; note that they are not the marginal effects carried by the links. When igraph=TRUE, this list is attached as the graph attribute coeffs.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

Examples

## Dobson (1990) Page 93: Randomized Controlled Trial :
counts <- c(18,17,15,20,10,20,25,13,12)
outcome <- gl(3,1,9)
treatment <- gl(3,3)
Dobson <- data.frame(counts=counts, outcome=outcome, treatment=treatment)
model <- "counts ~ outcome + treatment, poisson"
glmCoin(model,Dobson)

## Multinomial model: housing satisfaction in Copenhagen (Madsen, 1976)
data(housing, package="MASS")
g <- glmCoin("Sat ~ Infl + Type + Cont, multinomial", housing, weights="Freq")
g$equations

Description

produces a netCoin object from a set of glm regressions.

Usage

glmCoin2(formulas, data, weights=NULL, pmax=.05, 
                    twotail=FALSE, showArrows=TRUE,
                    frequency = FALSE, percentage = TRUE, 
                    color="variable", lwidth="z.value", 
                    circle= NA, language=c("en","es","ca"),
                    igraph=FALSE, ...)

Arguments

Value

This function creates a netCoin object (or igraph) and, if stated, a folder in the computer with an HTML document named index.html which contains the produced graph. This file can be directly opened with your browser and sent to a web server to work properly.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

Examples

## Dobson (1990) Page 93: Randomized Controlled Trial :
counts <- c(18,17,15,20,10,20,25,13,12)
outcome <- gl(3,1,9)
treatment <- gl(3,3)
Dobson <- data.frame(counts=counts, outcome=outcome, treatment=treatment)
model <- "counts ~ outcome + treatment, poisson"
glmCoin2(model,Dobson)

Description

Convert a data frame with two numbers (normally a beginning year and end year) into an incidences matrix whose rows are the intermediate numbers, and whose columns are the content of the names column.

Usage

incTime(data, name = "name", beginning = "birth", end= "death")

Arguments

Value

A data frame in which the two first columns are source and target. The rest of the columns are sim.=(1+threshold-real difference) and dist.=(difference between numbers)

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

Examples

# From sociologists data
data("sociologists")
head(incTime(sociologists))[,1:5]

Description

layoutCircle produces a circle layout of any number of nodes.

Usage

layoutCircle(N,nodes=seq_len(nrow(N)),deg=0,name=NULL)

Arguments

Value

This function returns the input data frame of nodes with the resulting layout applied.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

Examples

A <- data.frame(name=letters)

L <- layoutCircle(A,name="name")

netCoin(A,layout=L)

Description

layoutDist measures how far one and the same node travels from one plane of coordinates to another, added up over the nodes and divided by how many they are.

Usage

layoutDist(a, b, align=c("cloud", "procrustes", "canvas", "none"), weights=NULL,
           normalize=TRUE, who="layoutDist")

Arguments

Details

Every plane comes in its own units: the coordinates of a tsne reduction reach 60 where those of a classical one stop at 3, and nmf or a pair of observed variables live besides in the positive quadrant. Adding up the distances as they come measures little else than that difference of units, so the two planes are put on an equal footing first. How they are levelled decides what is being asked:

"cloud" centres every plane on its centroid and divides it by its root mean square radius. This strips translation and size away; what is left is a difference of configuration, orientation included. The divisor is one for the whole plane rather than one per column: standardising column by column would stretch the second axis until it matched the first, which undoes what a reduction states, that its first direction carries more variation than its second.

"procrustes" does the above and rotates besides the first plane, reflecting it where that helps, until it lies as close to the second as it can. It measures shape alone: two reductions ordering the cases alike but coming out rotated or mirrored end up at nearly no distance, which is the reasonable reading wherever the orientation of the axes means nothing, as in tsne or umap. No scaling factor of the fit is added, though the classical Procrustes statistic carries one: the sizes have been levelled already, and that factor would break the symmetry a matrix of distances between planes needs.

"canvas" brings every column to [0, 1] by its minimum and its maximum, which is what rD3plot does when it sets the domain of each axis to the range of the coordinates and its own range to the width or the height of the canvas. It measures the picture the reader is shown rather than the cloud behind it, and two planes differing only in the scale of their axes come out at zero. The canvas is taken square, as the shape of the window is not known from R.

"none" leaves the coordinates as they are, which fits where the units are already comparable, as between two pairs of variables measured from 0 to 100.

Rows holding NA in either plane are dropped, and the alignment is computed on the rows about to be compared rather than on the whole of them.

Value

A single number, the distance between the two planes. It carries as attributes the rows compared ("rows") and those dropped ("dropped"), the alignment applied ("align"), the divisor applied to each plane ("scales", one per plane under "cloud" and "procrustes", one per axis under "canvas") and, with align="procrustes" and planes of two columns, the rotation in degrees ("angle") and whether there was a reflection ("reflection").

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

See Also

layoutsDist, addAxes, surScat

Examples

## Not run: 
data(iris)
S <- surScat(iris, variables=names(iris)[1:4], type="pca", nclusters=3)
P <- prcomp(iris[,1:4], scale.=TRUE)$x
S <- addAxes(S, P[,3:4])
L <- S$layouts

layoutDist(L[[1]], L[[2]])                        # the cloud
layoutDist(L[[1]], L[[2]], align="procrustes")    # shape alone
layoutDist(L[[1]], L[[2]], align="canvas")        # the picture

z <- layoutDist(L[[1]], L[[2]], align="procrustes")
attr(z, "angle")
attr(z, "reflection")

## End(Not run)

Description

layoutGrid produces a grid layout of any number of nodes.

Usage

layoutGrid(N,string,name=NULL,byrow=FALSE)

Arguments

Value

This function returns the input data frame of nodes with the resulting layout applied.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

Examples

A <- data.frame(name=letters)

L <- layoutGrid(A,"a,b,c,d,e.f,g,h,i,j.k,l,m,n,o,p.q,r,s,t,u.v,w,x,y,z","name")

netCoin(A,layout=L)

Description

layoutsAlign brings the planes of a netCoin object from surScat to a common footing, as layoutsDist does to measure how far apart they are, and hands back the transformed coordinates themselves.

Usage

layoutsAlign(x, align=c("procrustes", "cloud", "canvas", "none"), to=1, gpa=FALSE,
             weights=NULL, which=NULL, suffix=NULL, tol=1e-10, maxit=100,
             who="layoutsAlign")

Arguments

Details

The fit is estimated on the rows that are complete in every plane involved, the same ones for all of them, and then applied to every row of each plane: a node missing from one plane keeps its place in the others.

The result is expressed in the units of the reference plane, which is therefore left just as it was (except under gpa=TRUE, where every plane moves towards the consensus). A plane compared with it under align="procrustes" is thus turned to be read on the same axes, and the distance layoutsDist reports between them is the mean distance between the two sets of coordinates divided by the scale of the reference.

Under gpa=TRUE no plane is favoured over the others. The mean has no orientation of its own, so the whole set is finally turned, as one, to lie as close as possible to the reference.

Value

When x is a 'netCoin' object, the same object with the aligned planes in place of the original ones in its $layouts (or each one next to its original, when suffix is given), the first plane also as its fx/fy columns, and the cluster centroids in $clusters computed again on the new coordinates. The details of the alignment go in its attribute "alignment".

When x is a list of planes, a named list holding the aligned ones, with those details as attributes of the list.

The details are: the method (align), the reference plane (to), whether gpa was used, the rows the fit was estimated on (rows) and those left out (dropped), the center, scale and rotation matrix applied to every plane, and, with align="procrustes" and two columns, the angle of each rotation in degrees and whether it involved a reflection. With gpa=TRUE there are also the consensus plane (the mean of all the aligned ones), the number of iterations and whether the fit converged.

Author(s)

Modesto Escobar, Department of Sociology and Communication, University of Salamanca.

See Also

layoutsDist, layoutDist, addAxes, surScat

Examples

## Not run: 
data(iris)
S <- surScat(iris, variables=names(iris)[1:4], type="pca", nclusters=3)
P <- prcomp(iris[,1:4], scale.=TRUE)$x
S <- addAxes(S, P[,3:4])
S <- addAxes(S, as.matrix(iris[,1:2]), name="sepal")

# every plane turned onto PCA, the one drawn first
A <- layoutsAlign(S)
A
attr(A, "alignment")$angle

# onto another plane, by its name
A <- layoutsAlign(S, to="sepal")

# the original planes kept, the aligned ones added as "PC3-PC4'" and "sepal'"
A <- layoutsAlign(S, suffix="'")
names(A$layouts)

# towards the mean of them all
G <- layoutsAlign(S, gpa=TRUE)
head(attr(G, "alignment")$consensus)

# the coordinates alone, from a list of planes
L <- layoutsAlign(S$layouts, which=c("sepal", "PC3-PC4"), to=1)

## End(Not run)

Description

layoutsDist measures how far apart every pair of planes of a netCoin object from surScat lies, and hands the result back as the dist object hclust, cmdscale and the like expect.

Usage

layoutsDist(x, align=c("cloud", "procrustes", "canvas", "none"), weights=NULL,
            which=NULL, normalize=TRUE, nperm=0,
            value=c("distance", "relative", "p", "chance"), digits=3, who="layoutsDist")

Arguments