gdverse

The goal of gdverse is to support the geodetector model and its variants.

Installation

You can install the development version of gdverse from GitHub with:

# install.packages("devtools")
devtools::install_github("SpatLyu/gdverse",build_vignettes = T,dep = T)

or install gdverse from r-universe:

install.packages('gdverse', repos='https://spatlyu.r-universe.dev')

Load data and package

library(sf)
library(terra)
library(tidyverse)
library(gdverse)
fvcpath = "https://github.com/SpatLyu/rdevdata/raw/main/FVC.tif"
fvc = terra::rast(paste0("/vsicurl/",fvcpath))
fvc = terra::aggregate(fvc,fact = 5)
fvc = as_tibble(terra::as.data.frame(fvc,na.rm = T))
head(fvc)
## # A tibble: 6 × 13
##     fvc premax premin presum tmpmax tmpmin tmpavg   pop   ntl  lulc  elev slope
##   <dbl>  <dbl>  <dbl>  <dbl>  <dbl>  <dbl>  <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 0.188   163.   6.86  3992.   21.2  -7.09   8.54  5.64  9.10    10 1645.  2.96
## 2 0.162   162.   5.23  3922.   21.7  -6.90   8.92 23.1  10.5     10 1539.  1.86
## 3 0.168   168.   4.15  4040.   21.2  -7.22   8.53  9.73  5.58    10 1611.  3.19
## 4 0.186   174.   5.99  4254.   20.8  -7.42   8.21  6.84  2.89    10 1677.  3.32
## 5 0.189   164.   7.86  4047.   21.2  -7.00   8.58  2.36 12.3     10 1643.  2.79
## 6 0.171   161.   5.23  3944.   21.7  -6.85   8.91  3.17 10.1     10 1553.  1.93
## # ℹ 1 more variable: aspect <dbl>

OPGD model

set.seed(12345678)
tictoc::tic()
fvc_opgd = opgd(fvc ~ ., data = fvc, discnum = 2:15,
                discvar = names(select(fvc,-c(fvc,lulc))),
                cores = 6, type = 'factor')
tictoc::toc()
## 3.08 sec elapsed

fvc_opgd
## Spatial Stratified Heterogeneity Test 
##  
##           Factor detector

variable	Q-statistic	P-value
presum	0.6642	6.083e-10
lulc	0.6597	8.782e-10
premin	0.4671	5.581e-10
tmpmin	0.4282	7.733e-10
tmpmax	0.2594	3.131e-10
elev	0.2353	2.885e-10
slope	0.2338	7.849e-10
tmpavg	0.2244	6.571e-10
pop	0.1953	2.428e-10
premax	0.1448	1.771e-10
ntl	0.02253	0.005407
aspect	0.0186	0.7491

RGD model

To run RGD,remember to set up your python dependence, see RGD vignette to get more details.

reticulate::use_condaenv('geocompy')
set.seed(12345678)
tictoc::tic()
fvc_rgd = rgd(fvc ~ ., data = fvc, discnum = 15, 
              discvar = names(select(fvc,-c(fvc,lulc))),
              cores = 6, type = c('factor','interaction'))
tictoc::toc()
## 1879.89 sec elapsed

fvc_rgd
## Spatial Stratified Heterogeneity Test 
##  
##           Factor detector

variable	Q-statistic	P-value
presum	0.6745	6.166e-10
lulc	0.6597	8.782e-10
premin	0.4855	5.461e-10
tmpmin	0.4575	6.232e-10
tmpmax	0.2816	5.853e-10
elev	0.2576	5.407e-10
pop	0.2566	4.386e-10
slope	0.2525	9.439e-10
tmpavg	0.2502	4.2e-10
premax	0.1668	9.434e-10
aspect	0.03427	0.06239
ntl	0.02334	4.623e-10