multigensd: Multiple Attributes skewed and dependent

In Sewanou/genspatattr: Generates skewed spatial attributes

Description

Generates multiple spatial dependent attributes with fixed skewness.

Usage

multigensd (n.attr, size, grid, sk = 1, dep = 0.5, mu = 0, v = 1,  method = "SAR", dst = "skewnormal")

Arguments

n.attr	Number of attributes to be generated.
size	a vector of two elements specifying respectively numbers of indivuals and time points of the attribute to be generated.
grid	number of rows and columns for grids.
sk	a scalar that represents the skewness coefficient to be considered.
dep	a scalar that represents the dependence correlation between values within a given attribute.
mu	the mean value
v	the variance value
method	a character string naming the the approach used to create spatial dependence between variables. The "SAR" method is the Kapoor, Kelejian, and Prucha (2007)-spatial autoregressive random effects (SAR-RE) model. The "SARMA" method represents the Lee and Yu (2012) spatial autoregressive moving average random effects (SARMA-RE) model.
dst	a character string naming the distribution to be used for vectors generation. "skewnormal" the skew-normal distribution is limited to a range of [0,1[ for the coefficient of skewness. "lognormal" the lognormal distribution allows a large range of coefficients skewness values.

Value

Returns a list.

mat.attributes	Multiple spatial dependent and skewed attributes generated.
spatial.dependence	Racall of the spatial dependence existing within values of a given attribute generated
skewness.coef	Coefficient of skewness of spatial attributes generated.

Author(s)

Sewanou Honfo <[e-mail]> and Brezesky Kotanmi <[e-mail]> and Eunice Gnanvi <[e-mail]> and Chenangnon Tovissode <[e-mail]> and Romain glèlè Kakaï <[e-mail]> / LABEF_08_2019

Examples

Consider a simulation study that requires 10 skewed (3.6) spatial attribbutes with a spatial autoregressive coefficient (spatial correlation) of 0.5.
Each attribute has 100 subjects and 7 observations (time points) per subject. We choose to design 10 x 10 square grids.

# Loading package
library(genspatattr)

# As the skewness required exceeds the interval [0,1], we use the lognormal distribution and SAR method to maintain null the spatial moving average coefficient.
k1 <- c(100, 7)
k2 <- c(10, 10)
dt <- multigensd(n.attr = 10, size = k1, grid = k2, sk = 3.6, dst = "lognormal")
> head(dt$mat.attributes)
       [,1]     [,2]      [,3]     [,4]      [,5]      [,6]      [,7]      [,8]      [,9]
[1,] 36.76639 29.65413 321.04957 54.14535  5.620346  38.27667 118.73450  44.93622  56.44374
[2,] 21.99757 27.46571  95.71541 37.93883  8.538977  82.19352  44.64520  39.25272  45.69996
[3,] 40.25177 32.81329  31.67804 34.86756 18.055213 384.21476  20.64373  42.84450  62.49243
[4,] 20.91775 50.49171  17.53125 35.79660 49.848041 171.73468  35.92800  44.00423 138.66885
[5,] 17.16706 37.16105  12.95207 50.99651 38.020209 159.57852  26.01867  75.07399  55.96092
[6,] 13.39583 35.46820  15.07685 36.63703 71.563533  43.86750  73.75986 173.56775  44.58139
      [,10]
[1,] 24.85834
[2,] 25.51857
[3,] 21.10237
[4,] 23.60568
[5,] 30.25069
[6,] 28.92886

> e1071::skewness(dt$mat.attributes[,1])
[1] 3.643574
> e1071::skewness(dt$mat.attributes[,10])
[1] 3.606651
> dt$spatial.dependence
[1] 0.5
> dt$skewness.coef
[1] 3.6  

Sewanou/genspatattr documentation built on Oct. 30, 2019, 11:52 p.m.