netdiffuseR (version 1.22.6)

survey_to_diffnet: Convert survey-like data and edgelists to a diffnet object

Description

These convenient functions turn network nomination datasets and edgelists with vertex attributes datasets into diffnet objects. Both work as wrappers of edgelist_to_adjmat and new_diffnet.

Usage

survey_to_diffnet(
  dat,
  idvar,
  netvars,
  toavar,
  groupvar = NULL,
  no.unsurveyed = TRUE,
  timevar = NULL,
  t = NULL,
  undirected = getOption("diffnet.undirected", FALSE),
  self = getOption("diffnet.self", FALSE),
  multiple = getOption("diffnet.multiple", FALSE),
  keep.isolates = TRUE,
  recode.ids = TRUE,
  warn.coercion = TRUE,
  ...
)
edgelist_to_diffnet(
  edgelist,
  w = NULL,
  t0 = NULL,
  t1 = NULL,
  dat,
  idvar,
  toavar,
  timevar = NULL,
  undirected = getOption("diffnet.undirected", FALSE),
  self = getOption("diffnet.self", FALSE),
  multiple = getOption("diffnet.multiple", FALSE),
  fill.missing = NULL,
  keep.isolates = TRUE,
  recode.ids = TRUE,
  warn.coercion = TRUE
)

Value

A diffnet object.

Arguments

dat

A data frame.

idvar

Character scalar. Name of the id variable.

netvars

Character vector. Names of the network nomination variables.

toavar

Character scalar. Name of the time of adoption variable.

groupvar

Character scalar. Name of cohort variable (e.g. city).

no.unsurveyed

Logical scalar. When TRUE the nominated individuals that do not show in idvar are set to NA (see details).

timevar

Character sacalar. In the case of longitudinal data, name of the time var.

t

Integer scalar. Repeat the network t times (if no t0,t1 are provided).

undirected

Logical scalar. When TRUE only the lower triangle of the adjacency matrix will considered (faster).

self

Logical scalar. When TRUE autolinks (loops, self edges) are allowed (see details).

multiple

Logical scalar. When TRUE allows multiple edges.

keep.isolates

Logical scalar. When FALSE, rows with NA/NULL values (isolated vertices unless have autolink) will be droped (see details).

recode.ids

Logical scalar. When TRUE ids are recoded using as.factor (see details).

warn.coercion

Logical scalar. When TRUE warns coercion from numeric to integer.

...

Further arguments to be passed to new_diffnet.

edgelist

Two column matrix/data.frame in the form of ego -source- and alter -target- (see details).

w

Numeric vector. Strength of ties (optional).

t0

Integer vector. Starting time of the ties (optional).

t1

Integer vector. Finishing time of the ties (optional).

fill.missing

Character scalar. In the case of having unmatching ids between dat and edgelist, fills the data (see details).

Author

Vega Yon

Details

All of netvars, toavar and groupvar must be integers. Were these numeric they are coerced into integers, otherwise, when neither of both, the function returns with error. idvar, on the other hand, should only be integer when calling survey_to_diffnet, on the contrary, for edgelist_to_diffnet, idvar may be character.

In field work it is not unusual that some respondents nominate unsurveyed individuals. In such case, in order to exclude them from the analysis, the user can set no.unsurveyed=TRUE (the default), telling the function to exclude such individuals from the adjacency matrix. This is done by setting variables in netvars equal to NA when the nominated id can't be found in idvar.

If the network nomination process was done in different groups (location for example) the survey id numbers may be define uniquely within each group but not across groups (there may be many individuals with id=1, for example). To encompass this issue, the user can tell the function what variable can be used to distinguish between groups through the groupvar argument. When groupvar is provided, function redifines idvar and the variables in netvars as follows:

   dat[[idvar]] <- dat[[idvar]] + dat[[groupvar]]*z

Where z = 10^nchar(max(dat[[idvar]])).

For longitudinal data, it is assumed that the toavar holds the same information through time, this is, time-invariable. This as the package does not yet support variable times of adoption.

The fill.missing option can take any of these three values: "edgelist", "dat", or "both". This argument works as follows:

1. When fill.missing="edgelist" (or "both") the function will check which vertices show in dat but do not show in edgelist. If there is any, the function will include these in edgelist as ego to NA (so they have no link to anyone), and, if specified, will fill the t0, t1 vectors with NAs for those cases. If w is also specified, the new vertices will be set to min(w, na.rm=TRUE).

2. When fill.missing="dat" (or "both") the function checks which vertices show in edgelist but not in dat. If there is any, the function will include these in dat by adding one row per individual.

See Also

fakesurvey, fakesurveyDyn

Other data management functions: diffnet-class, edgelist_to_adjmat(), egonet_attrs(), isolated()

Examples

# Loading a fake survey (data frame)
data(fakesurvey)

# Diffnet object keeping isolated vertices ----------------------------------
dn1 <- survey_to_diffnet(fakesurvey, "id", c("net1", "net2", "net3"), "toa",
   "group", keep.isolates=TRUE)

# Diffnet object NOT keeping isolated vertices
dn2 <- survey_to_diffnet(fakesurvey, "id", c("net1", "net2", "net3"), "toa",
   "group", keep.isolates=FALSE)

# dn1 has an extra vertex than dn2
dn1
dn2

# Loading a longitudinal survey data (two waves) ----------------------------
data(fakesurveyDyn)

groupvar <- "group"
x <- survey_to_diffnet(
   fakesurveyDyn, "id", c("net1", "net2", "net3"), "toa", "group" ,
   timevar = "time", keep.isolates = TRUE, warn.coercion=FALSE)

plot_diffnet(x, vertex.label = rownames(x))

# Reproducing medInnovationsDiffNet object ----------------------------------
data(medInnovations)

# What are the netvars
netvars <- names(medInnovations)[grepl("^net", names(medInnovations))]

medInnovationsDiffNet2 <- survey_to_diffnet(
   medInnovations,
   "id", netvars, "toa", "city",
   warn.coercion=FALSE)

medInnovationsDiffNet2

# Comparing with the package's version
all(diffnet.toa(medInnovationsDiffNet2) == diffnet.toa(medInnovationsDiffNet)) #TRUE
all(
   diffnet.attrs(medInnovationsDiffNet2, as.df = TRUE) ==
   diffnet.attrs(medInnovationsDiffNet, as.df = TRUE),
   na.rm=TRUE) #TRUE