randomizations {spatsoc}

Data-stream randomizations

Description

randomizations performs data-stream social network randomization. The function accepts a data.table with relocation data, individual identifiers and a randomization type. The data.table is randomized either using step or daily between-individual methods, or within-individual daily trajectory method described by Spiegel et al. (2016).

Usage

randomizations(
  DT = NULL,
  type = NULL,
  id = NULL,
  group = NULL,
  coords = NULL,
  datetime = NULL,
  splitBy = NULL,
  iterations = NULL
)

Arguments

DT

input data.table

type

one of 'daily', 'step' or 'trajectory' - see details

id

Character string of ID column name

group

generated from spatial grouping functions - see details

coords

Character vector of X coordinate and Y coordinate column names

datetime

field used for providing date time or time group - see details

splitBy

List of fields in DT to split the randomization process by

iterations

The number of iterations to randomize

Details

The DT must be a data.table. If your data is a data.frame, you can convert it by reference using data.table::setDT.

Three randomization types are provided:

1. step - randomizes identities of relocations between individuals within each time step.

2. daily - randomizes identities of relocations between individuals within each day.

3. trajectory - randomizes daily trajectories within individuals (Spiegel et al. 2016).

Depending on the type, the datetime must be a certain format:

• step - datetime is integer group created by group_times

• daily - datetime is POSIXct format

• trajectory - datetime is POSIXct format

The id, datetime, (and optional splitBy) arguments expect the names of respective columns in DT which correspond to the individual identifier, date time, and additional grouping columns. The coords argument is only required when the type is "trajectory", since the coordinates are required for recalculating spatial groups with group_pts, group_lines or group_polys.

Please note that if the data extends over multiple years, a column indicating the year should be provided to the splitBy argument. This will ensure randomizations only occur within each year.

The group argument is expected only when type is 'step' or 'daily'.

For example, using data.table::year:

 DT[, yr := year(datetime)] randomizations(DT, type = 'step',
id = 'ID', datetime = 'timegroup', splitBy = 'yr') 

iterations is set to 1 if not provided. Take caution with a large value for iterations with large input DT.

Value

randomizations returns the random date time or random id along with the original DT, depending on the randomization type. The length of the returned data.table is the original number of rows multiplied by the number of iterations + 1. For example, 3 iterations will return 4x - one observed and three randomized.

Two columns are always returned:

• observed - if the rows represent the observed (TRUE/FALSE)

• iteration - iteration of rows (where 0 is the observed)

In addition, depending on the randomization type, random ID or random date time columns are returned:

• step - randomID each time step

• daily - randomID for each day and jul indicating julian day

• trajectory - a random date time ("random" prefixed to datetime argument), observed jul and randomJul indicating the random day relocations are swapped to.

References

doi:10.1111/2041-210X.12553

See Also

Other Social network tools: get_gbi()

Examples

# Load data.table
library(data.table)


# Read example data
DT <- fread(system.file("extdata", "DT.csv", package = "spatsoc"))

# Select only individuals A, B, C for this example
DT <- DT[ID %in% c('A', 'B', 'C')]

# Date time columns
DT[, datetime := as.POSIXct(datetime)]
DT[, yr := year(datetime)]

# Temporal grouping
group_times(DT, datetime = 'datetime', threshold = '5 minutes')

# Spatial grouping with timegroup
group_pts(DT, threshold = 5, id = 'ID', coords = c('X', 'Y'), timegroup = 'timegroup')

# Randomization: step
randStep <- randomizations(
    DT,
    type = 'step',
    id = 'ID',
    group = 'group',
    datetime = 'timegroup',
    splitBy = 'yr',
    iterations = 2
)

# Randomization: daily
randDaily <- randomizations(
    DT,
    type = 'daily',
    id = 'ID',
    group = 'group',
    datetime = 'datetime',
    splitBy = 'yr',
    iterations = 2
)

# Randomization: trajectory
randTraj <- randomizations(
    DT,
    type = 'trajectory',
    id = 'ID',
    group = NULL,
    coords = c('X', 'Y'),
    datetime = 'datetime',
    splitBy = 'yr',
    iterations = 2
)