sienaDependent {RSiena}

Function to create a dependent variable for a Siena model

Description

Creates a Siena dependent variable: either a network, created from a matrix or array or list of sparse matrix of triples; or a behavior variable, created from a matrix.
sienaDependent() and sienaNet() are identical functions; the second name was used from the start of the RSiena package, but the first name indicates more precisely the purpose of this function.

Usage

sienaDependent(netarray, type=c("oneMode", "bipartite", "behavior", "continuous"),
nodeSet="Actors", sparse=is.list(netarray), allowOnly=TRUE, imputationValues=NULL)

sienaNet(netarray, type=c("oneMode", "bipartite", "behavior", "continuous"),
nodeSet="Actors", sparse=is.list(netarray), allowOnly=TRUE, imputationValues=NULL)

Arguments

netarray

type="behavior" or "continuous": matrix (actors \times waves).
type="oneMode" or "bipartite": array of values or list of sparse matrices of type "TsparseMatrix", see the Matrix package; if an array is used, it should have dimensions: for a one-mode network, n \times n \times M, and for a two-mode network n \times m \times M, where n is the number of actors, m is the number of nodes in the second mode, and M is the number of waves.

type

type of dependent variable, default oneMode.

nodeSet

character string naming the appropriate node set. For a bipartite network, a vector containing 2 character strings: "rows" first, then "columns".

sparse

logical: TRUE indicates the data is in sparse matrix format, FALSE otherwise.

allowOnly

logical: If TRUE, it will be detected when between any two consecutive waves the changes are non-decreasing or non-increasing, and if this is the case, this will also be a constraint for the simulations between these two waves. This is done by means of the internal parameters uponly and downonly. If FALSE, the parameters uponly and downonly always are set to FALSE, and changes in dependent variables will not be constrained to be non-decreasing or non-increasing. This also will imply that some effects are excluded because they are superfluous in such constrained situations. This will be reported in the output of print01Report.

For normal operation when this is the case for all periods, usually TRUE is the appropriate option. When it is only the case for some of the periods, and for data sets that will be part of a multi-group object created by sienaGroupCreate, FALSE usually is preferable.

imputationValues

for behavior or continuous dependent variables, a matrix with imputation values can be included that will be used instead of the default imputation values.

Details

Adds attributes so that the array or list of matrices can be used in a Siena model fit.

Value

An object of class sienaDependent. An array or (networks only) a list of sparse matrices with attributes:

netdims

Dimensions of the network or behavior variable: senders, receivers (1 for behavior), periods

type

oneMode, bipartite or behavior

sparse

Boolean: whether the network is given as a list of sparse matrices or not

nodeSet

Character string with name(s) of node set(s)

allowOnly

The value of the allowOnly parameter

Author(s)

Ruth Ripley and Tom A.B. Snijders

References

See https://www.stats.ox.ac.uk/~snijders/siena/ .

See Also

sienaDataCreate, sienaNodeSet, sienaDataConstraint

Examples

mynet1 <- sienaDependent(array(c(s501, s502, s503), dim=c(50, 50, 3)))
mybeh <- sienaDependent(s50a, type="behavior")
## note that the following example works although the node sets do not yet exist!
mynet3 <- sienaDependent(array(c(s501, s502, s503), dim=c(50, 50, 3)),
       type="bipartite", nodeSet=c("senders", "receivers"))
## sparse matrix input
## To show this, we first go back from the adjacency matrices to edgelists.
## The manual shows one way to do this.
## Another way is to use the sparse matrix representation which internally
## indeed is an edge list:
library(Matrix)
sp501 <- as(Matrix(s501), "TsparseMatrix")
sp502 <- as(Matrix(s502), "TsparseMatrix")
sp503 <- as(Matrix(s503), "TsparseMatrix")
## If you are interested in the internal structure of these sparse matrices,
## you can request
str(sp501)
## Slot @i is the row, @j is the column, and @x the value;
## here the values all are 1.
## Slots @i and @j do not contain information about the number of nodes,
## so that is supplied additionally by @Dim.
mymatlist <- list(sp501, sp502, sp503)
mynet.sp <- sienaDependent(mymatlist)