graph4lg (version 1.8.0)

dist_max_corr: Compute the distance at which the correlation between genetic distance and landscape distance is maximal

Description

The function enables to compute the distance at which the correlation between genetic distance and landscape distance is maximal, using a method similar to that employed by van Strien et al. (2015). Iteratively, distance threshold values are tested. For each value, all the population pairs separated by a landscape distance larger than the threshold are removed before the Mantel correlation coefficient between genetic distance and landscape distance is computed. The distance threshold at which the correlation is the strongest is then identified. A figure showing the evolution of the correlation coefficients when landscape distance threshold increases is plotted.

Usage

dist_max_corr(
  mat_gd,
  mat_ld,
  interv,
  from = NULL,
  to = NULL,
  fig = TRUE,
  thr_gd = NULL,
  line_col = "black",
  pts_col = "#999999"
)

Value

A list of objects:

• The distance at which the correlation is the highest.

• The vector of correlation coefficients at the different distance thresholds

• The vector of the different distance thresholds

• A ggplot2 object to plot

Arguments

mat_gd

A symmetric matrix or dist object with pairwise genetic distances between populations or sample sites.

mat_ld

A symmetric matrix or dist object with pairwise landscape distances between populations or sample sites. These distances can be Euclidean distances, cost-distances or resistance distances, among others.

interv

A numeric or integer value indicating the interval between the different distance thresholds for which the correlation coefficients are computed.

from

(optional) The minimum distance threshold value at which the correlation coefficient is computed.

to

(optional) The maximum distance threshold value at which the correlation coefficient is computed.

fig

Logical (default = TRUE) indicating whether a figure is plotted.

thr_gd

(optional) A numeric or integer value used to remove genetic distance values from the data before the calculation. All genetic distances values above 'thr_gd' are removed from the data. This parameter can be used especially when there are outliers.

line_col

(optional, if fig = TRUE) A character string indicating the color used to plot the line (default: "blue"). It must be a hexadecimal color code or a color used by default in R.

pts_col

(optional, if fig = TRUE) A character string indicating the color used to plot the points (default: "#999999"). It must be a hexadecimal color code or a color used by default in R.

Author

P. Savary

Details

IDs in 'mat_gd' and 'mat_ld' must be the same and refer to the same sampling sites or populations, and both matrices must be ordered in the same way. The correlation coefficient between genetic distance and landscape distance computed is a Mantel correlation coefficient. If there are less than 50 pairwise values, the correlation is not computed, as in van Strien et al. (2015). Such a method can be subject to criticism from a strict statistical point of view given correlation coefficients computed from samples of different size are compared. The matrix of genetic distance 'mat_gd' can be computed using mat_gen_dist. The matrix of landscape distance 'mat_ld' can be computed using mat_geo_dist when the landscape distance needed is a Euclidean geographical distance. Mantel correlation coefficients are computed using the function mantel.

References

van2015isolationgraph4lg

Examples

data("data_tuto")
mat_gen <- data_tuto[[1]]
mat_dist <- data_tuto[[2]]*1000
res_dmc <- dist_max_corr(mat_gd = mat_gen,
                         mat_ld = mat_dist,
                         from = 32000, to = 42000,
                         interv = 5000,
                         fig = FALSE)