Simulates intergenerational dispersal in a species defined by multiple dispersal components across the breeding cycle, with
dispersal, breeding & sampling & basic generational structure custom-defined by a DispersalModel object.
simulate_kindist_custom(
nsims = 100,
model = dispersal_model(init = 100, breed = 50, grav = 50, ovi = 25, .FS = "ovi", .HS
= "breed"),
dims = 100,
method = "Gaussian",
kinship = "FS",
cycle = 0,
shape = 0.5
)(integer) - number of pairs to simulate
(object of class DispersalModel) - species-specific model of dispersal containing lifestage, phase & sampling parameters
(numeric) - length of sides of (square) simulated site area
(character) - kernel shape to use: either 'Gaussian', 'Laplace' or 'vgamma' (variance-gamma)
(character)- kin category to simulate: one of PO, FS, HS, AV, GG, HAV, GGG, 1C, 1C1, 2C, GAV, HGAV, H1C H1C1 or H2C
(numeric) - breeding cycle number(s) of dispersed kin to be modeled. Must be a integer equal to or greater than -1, (-1, 0, 1, 2, ...)
or vector of two such integers. Represents the number of complete breeding cycles each simulated individual has undergone before the sampling point,
where the time between birth and first reproduction is coded as '0', that between first and second reproduction '1', etc. (default 0).
If cycle is specially set to '-1' this constitutes the sampling of an individual before it has differentiated (via dispersal) from the parent.
Only use in spp. where there is likely to be a reasonable equivalence between breeding stages across a lifespan.
(numeric) - value of shape parameter to use with 'vgamma' method. Default 0.5. Must be > 0. Increment towards zero for increasingly heavy-tailed (leptokurtic) dispersal
returns an object of class KinPairSimulation containing simulation details and a tibble (tab) of simulation values
This function is one of a family of functions that implement the core intergenerational dispersal simulations
contained in the kindisperse package. Each of these functions proceeds by the following steps:
identify the pedigree relationship, dispersal phase (FS, HS & PO) and sampling stage that must be generated;
randomly assign a coordinate position to the 'root' individual within the pedigree (i.e. last common ancestor of the dyad, inclusive);
'disperse' both pathways from this root position via the appropriately defined phase dispersal (additively via random draws from the underlying statistical model, defined by an axial standard deviation - sigma);
further disperse both phased descendant branches according to the number of realised breeding dispersal cycles contained in the defining pedigree (additively via random draws from the chosen underlying statistical model);
add displacement caused by dispersal before the sampling point in a similar manner to above, defining the final positions of the sampled dispersed kin dyads;
calculating geographical distances between the resulting dyads.
These simulation functions operate under an additive variance framework: all individual dispersal events are modeled as random
draws from a bivariate probability distribution defined by an axial standard deviation sigma and (sometimes) a shape
parameter. At present, three such distributions are included as options accessible with the method parameter: the
bivariate normal distribution 'Gaussian', the bivariate Laplace distribution 'Laplace', and the bivariate
variance-gamma distribution 'vgamma'. The Gaussian (normal) distribution enables easy compatibility with the
framework under which much population genetic & dispersal theory (isolation by distance, neighbourhoods, etc.) have been
developed. The Laplace distribution is a multivariate adaptation of the (positive) exponential distribution, and
represents a more 'fat-tailed' (leptokurtic) disperal situation than Gaussian. The vgamma distribution is a mixture
distribution formed by mixing the gamma distribution with the bivariate normal distribution. The flexibility of this
distribution's shape parameter enables us to model arbitrarily leptokurtic dispesal kernels, providing a helpful way
to examine the impacts of (e.g.) long distance dispersal on the overall disperal distribution and sampling decisions. A
vgamma distribution with shape parameter equal to 1 reduces to the bivariate Laplace distribution. As shape approaches
infinity, the vgamma distribution approaches the bivariate normal distribution. As shape approaches zero, the distribution
becomes increasingly leptokurtic.
The simulate_kindist_custom() function is designed to enable modeling of the composite dispersal events that occur
within the breeding cycle of an organism, and enables the separate treatment of the PO, FS, and HS
phases in situations where the breeding and dispersal cycle of an organism is (somewhat more complex that that
encountered in organisms such as mosquitoes (i.e. single-generational breeding organisms with defined sampling
points). This function relies on a custom dispersal model of class DispersalModel defined via parameter
model to supply organism-specific information about dispersal stages (with axial sigmas), FS & HS branch points, and
the dispersal stage at which sampling occurs. Via this model object (or overridden by the cycle parameter) you can
also define the number of breeding cycles each final individual within the close-kin dyad has passed through before sampling.
This is defined as a length one or two non-negative integer (where a length-one integer of value a is converted to a length
two integer of value c(a, a)), where the first integer defines the number of life cycles passed through by the 'senior'
pedigree member of the dyad, and the second the number passed through by the 'junior' member (so the GG phase has a grandparent
as senior, the grandchild as junior, etc. (in practice this distinction is unimportant for dyads). A cycle number of 0
references an individual that hasn't lived through an entire breeding cycle (sampling phase to sampling phase) before being
sampled. A value of 1 references an individual that has lived through one such cycle (e.g. a female entering her second breeding
season, an ovipositing mosquito (where the oviposition dispersal stage overlaps with the larval dispersal stage)). A value of
2 references two such cycles, etc. As all cycles are considered equivalent in the current formulation of this model (whether
an individual enters the cycle as a juvenile or as an adult) care must be taken in applying this system to species where the
dispersal behaviour of a second cycle individual (i.e. adult) is likely to be substantially different to that of a first cycle
individual (often an immature individual).
This function can only handle one kinship pairing & dispersal mode in the one simulation: where multiple dispersal pathways lead to the same kinship outcome, each pathway should be simulated separately, and the resulting distributions combined subsequently.
Following simulation, the results are returned as an object of the specially defined package class KinPairSimulation,
which stores the simulation results along with information about all simulation parameters, and can be further passed to
sample filtering & dispersal estimation functions.
Other simulate_kindist:
simulate_kindist_composite(),
simulate_kindist_simple()
# NOT RUN {
custom_dispersal_model <- dispersal_model(a = 10, b = 25, .FS = "b",
.HS = "a", .sampling_stage = "b")
simulate_kindist_custom(nsims = 100, model = custom_dispersal_model,
cycle = c(0, 1), kinship = "FS")
# }
Run the code above in your browser using DataLab