The hardware and bandwidth for this mirror is donated by METANET, the Webhosting and Full Service-Cloud Provider.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]metanet.ch.

Geographical Detector(GD)

This vignette explains how to run native geographic detector(GD) in gdverse package.

Load package and pre-processing data.

library(sf)
library(tidyverse)
library(gdverse)

See layers in NTDs.gpkg:

ntdspath = system.file("extdata", "NTDs.gpkg",package = 'gdverse')
st_layers(ntdspath)
## Driver: GPKG 
## Available layers:
##   layer_name geometry_type features fields crs_name
## 1    disease       Polygon      189      2  unknown
## 2  watershed       Polygon        9      2  unknown
## 3  elevation Multi Polygon        7      2  unknown
## 4   soiltype Multi Polygon        6      2  unknown

In file NTDs.gpkg, disease layer is the dependent variable, which is a continuous numerical variable, while others are independent and discrete variables.

Now we need to combine these layers together:

watershed = read_sf(ntdspath,layer = 'watershed')
elevation = read_sf(ntdspath,layer = 'elevation')
soiltype = read_sf(ntdspath,layer = 'soiltype')
disease = read_sf(ntdspath,layer = 'disease')

Plot them together:

library(cowplot)

f1 = ggplot(data = disease) +
  geom_sf(aes(fill = incidence),lwd = .1,color = 'grey') +
  viridis::scale_fill_viridis(option="mako", direction = -1) +
  theme_bw() +
  theme(
    axis.text = element_blank(),
    axis.ticks = element_blank(),
    axis.title = element_blank(),
    panel.grid = element_blank(),
    legend.position = 'inside',
    legend.position.inside = c(.1,.25),
    legend.background = element_rect(fill = 'transparent',color = NA)
  )
f2 = ggplot(data = watershed) +
  geom_sf(aes(fill = watershed),lwd = .1,color = 'grey') +
  tidyterra::scale_fill_whitebox_c() +
  coord_sf(crs = NULL) +
  theme_bw() +
  theme(
    axis.text = element_blank(),
    axis.ticks = element_blank(),
    axis.title = element_blank(),
    panel.grid = element_blank(),
    legend.position = 'inside',
    legend.position.inside = c(.1,.25),
    legend.background = element_rect(fill = 'transparent',color = NA)
  )
f3 = ggplot(data = elevation) +
  geom_sf(aes(fill = elevation),lwd = .1,color = 'grey') +
  tidyterra::scale_fill_hypso_c() +
  theme_bw() +
  theme(
    axis.text = element_blank(),
    axis.ticks = element_blank(),
    axis.title = element_blank(),
    panel.grid = element_blank(),
    legend.position = 'inside',
    legend.position.inside = c(.1,.25),
    legend.background = element_rect(fill = 'transparent',color = NA)
  )
f4 = ggplot(data = soiltype) +
  geom_sf(aes(fill = soiltype),lwd = .1,color = 'grey') +
  tidyterra::scale_fill_wiki_c() +
  theme_bw() +
  theme(
    axis.text = element_blank(),
    axis.ticks = element_blank(),
    axis.title = element_blank(),
    panel.grid = element_blank(),
    legend.position = 'inside',
    legend.position.inside = c(.1,.25),
    legend.background = element_rect(fill = 'transparent',color = NA)
  )

plot_grid(f1,f2,f3,f4, nrow = 2,label_fontfamily = 'serif',
          labels = paste0('(',letters[1:4],')'),
          label_fontface = 'plain',label_size = 10,
          hjust = -1.5,align = 'hv')  -> p
p

Attribute spatial join

NTDs = disease %>%
  st_centroid() %>%
  st_join(watershed[,"watershed"]) %>%
  st_join(elevation[,"elevation"]) %>%
  st_join(soiltype[,"soiltype"])

Check whether has NA in NTDs:

NTDs %>%
  dplyr::filter(if_any(everything(),~is.na(.x)))
## Simple feature collection with 4 features and 5 fields
## Geometry type: POINT
## Dimension:     XY
## Bounding box:  xmin: 301567.5 ymin: 3989433 xmax: 318763.3 ymax: 3991906
## Projected CRS: +proj=aea +lat_0=0 +lon_0=105 +lat_1=25 +lat_2=47 +x_0=0 +y_0=0 +datum=WGS84 +units=m +no_defs
## # A tibble: 4 × 6
##   SP_ID incidence               geom watershed elevation soiltype
## * <chr>     <dbl>        <POINT [m]>     <int>     <int>    <int>
## 1 141        6.48 (318763.3 3991847)         9        NA        3
## 2 165        6.53 (316574.1 3989433)         4        NA        2
## 3 166        6.43 (311439.1 3990674)         4        NA        2
## 4 188        6.26 (301567.5 3991906)        NA         2        3

Filter out all rows with no NA values:

NTDs %>%
  dplyr::filter(if_all(everything(),~!is.na(.x))) -> NTDs
NTDs
## Simple feature collection with 185 features and 5 fields
## Geometry type: POINT
## Dimension:     XY
## Bounding box:  xmin: 272784.9 ymin: 3982369 xmax: 333088.5 ymax: 4025117
## Projected CRS: +proj=aea +lat_0=0 +lon_0=105 +lat_1=25 +lat_2=47 +x_0=0 +y_0=0 +datum=WGS84 +units=m +no_defs
## # A tibble: 185 × 6
##    SP_ID incidence               geom watershed elevation soiltype
##  * <chr>     <dbl>        <POINT [m]>     <int>     <int>    <int>
##  1 0          5.94 (287216.1 4004408)         5         5        5
##  2 1          5.87 (284861.2 4001734)         5         5        4
##  3 2          5.88 (287731.3 4000289)         5         5        5
##  4 3          5.98 (290808.8 4000994)         5         5        5
##  5 4          5.96 (290924.8 4004432)         5         5        1
##  6 5          5.66 (272784.9 4007758)         5         5        4
##  7 6          5.74 (274308.9 4010571)         5         5        4
##  8 7          5.88 (279539.7 4003222)         5         5        4
##  9 8          6.1  (279032.9 4005677)         5         5        4
## 10 9          5.89 (282534.9 4004849)         5         5        4
## # ℹ 175 more rows

Remove unnecessary columns of data:

NTDs = NTDs %>%
  st_drop_geometry() %>%
  dplyr::select(-SP_ID)
NTDs
## # A tibble: 185 × 4
##    incidence watershed elevation soiltype
##        <dbl>     <int>     <int>    <int>
##  1      5.94         5         5        5
##  2      5.87         5         5        4
##  3      5.88         5         5        5
##  4      5.98         5         5        5
##  5      5.96         5         5        1
##  6      5.66         5         5        4
##  7      5.74         5         5        4
##  8      5.88         5         5        4
##  9      6.1          5         5        4
## 10      5.89         5         5        4
## # ℹ 175 more rows

Factor detector

fd = gd(incidence ~ watershed + elevation + soiltype,
        data = NTDs,type = 'factor')
fd
##                 Factor Detector            
## 
## | variable  | Q-statistic |   P-value   |
## |:---------:|:-----------:|:-----------:|
## | watershed |  0.6377737  | 0.000128803 |
## | elevation |  0.6067087  | 0.043382244 |
## | soiltype  |  0.3857168  | 0.372145486 |
plot(fd)

Interaction detector

id = gd(incidence ~ watershed + elevation + soiltype,
        data = NTDs,type = 'interaction')
id
##                 Interaction Detector         
## 
## | Interactive variable  | Interaction  |
## |:---------------------:|:------------:|
## | watershed ∩ elevation | Enhance, bi- |
## | watershed ∩ soiltype  | Enhance, bi- |
## | elevation ∩ soiltype  | Enhance, bi- |
plot(id)

Risk detector

rd = gd(incidence ~ watershed + elevation + soiltype,
        data = NTDs,type = 'risk')
rd
##                 Risk Detector            
## 
##  Variable elevation:
## 
## | zone  | zone1 | zone2 | zone3 | zone4 | zone5 | zone6 |
## |:-----:|:-----:|:-----:|:-----:|:-----:|:-----:|:-----:|
## | zone2 |  Yes  |  NA   |  NA   |  NA   |  NA   |  NA   |
## | zone3 |  Yes  |  No   |  NA   |  NA   |  NA   |  NA   |
## | zone4 |  Yes  |  Yes  |  Yes  |  NA   |  NA   |  NA   |
## | zone5 |  Yes  |  Yes  |  Yes  |  Yes  |  NA   |  NA   |
## | zone6 |  Yes  |  Yes  |  Yes  |  Yes  |  Yes  |  NA   |
## | zone7 |  Yes  |  Yes  |  Yes  |  Yes  |  Yes  |  Yes  |
## 
##  Variable soiltype:
## 
## | zone  | zone1 | zone2 | zone3 | zone4 |
## |:-----:|:-----:|:-----:|:-----:|:-----:|
## | zone2 |  Yes  |  NA   |  NA   |  NA   |
## | zone3 |  Yes  |  No   |  NA   |  NA   |
## | zone4 |  Yes  |  Yes  |  Yes  |  NA   |
## | zone5 |  No   |  Yes  |  Yes  |  Yes  |
## 
##  Variable watershed:
## 
## | zone  | zone1 | zone2 | zone3 | zone4 | zone5 | zone6 | zone7 | zone8 |
## |:-----:|:-----:|:-----:|:-----:|:-----:|:-----:|:-----:|:-----:|:-----:|
## | zone2 |  Yes  |  NA   |  NA   |  NA   |  NA   |  NA   |  NA   |  NA   |
## | zone3 |  Yes  |  Yes  |  NA   |  NA   |  NA   |  NA   |  NA   |  NA   |
## | zone4 |  Yes  |  No   |  Yes  |  NA   |  NA   |  NA   |  NA   |  NA   |
## | zone5 |  Yes  |  Yes  |  Yes  |  Yes  |  NA   |  NA   |  NA   |  NA   |
## | zone6 |  Yes  |  Yes  |  Yes  |  Yes  |  No   |  NA   |  NA   |  NA   |
## | zone7 |  Yes  |  Yes  |  No   |  Yes  |  Yes  |  Yes  |  NA   |  NA   |
## | zone8 |  Yes  |  Yes  |  No   |  Yes  |  Yes  |  Yes  |  No   |  NA   |
## | zone9 |  Yes  |  Yes  |  No   |  Yes  |  Yes  |  Yes  |  No   |  Yes  |
plot(rd)

You can change the significant interval by assign alpha argument,the default value of alpha argument is 0.95.

rd99 = gd(incidence ~ watershed + elevation + soiltype,
          data = NTDs,type = 'risk',alpha = 0.99)
rd99
##                 Risk Detector            
## 
##  Variable elevation:
## 
## | zone  | zone1 | zone2 | zone3 | zone4 | zone5 | zone6 |
## |:-----:|:-----:|:-----:|:-----:|:-----:|:-----:|:-----:|
## | zone2 |  Yes  |  NA   |  NA   |  NA   |  NA   |  NA   |
## | zone3 |  Yes  |  No   |  NA   |  NA   |  NA   |  NA   |
## | zone4 |  Yes  |  Yes  |  Yes  |  NA   |  NA   |  NA   |
## | zone5 |  Yes  |  Yes  |  Yes  |  Yes  |  NA   |  NA   |
## | zone6 |  Yes  |  Yes  |  Yes  |  Yes  |  Yes  |  NA   |
## | zone7 |  Yes  |  Yes  |  Yes  |  Yes  |  Yes  |  Yes  |
## 
##  Variable soiltype:
## 
## | zone  | zone1 | zone2 | zone3 | zone4 |
## |:-----:|:-----:|:-----:|:-----:|:-----:|
## | zone2 |  Yes  |  NA   |  NA   |  NA   |
## | zone3 |  Yes  |  No   |  NA   |  NA   |
## | zone4 |  Yes  |  Yes  |  Yes  |  NA   |
## | zone5 |  No   |  Yes  |  Yes  |  Yes  |
## 
##  Variable watershed:
## 
## | zone  | zone1 | zone2 | zone3 | zone4 | zone5 | zone6 | zone7 | zone8 |
## |:-----:|:-----:|:-----:|:-----:|:-----:|:-----:|:-----:|:-----:|:-----:|
## | zone2 |  Yes  |  NA   |  NA   |  NA   |  NA   |  NA   |  NA   |  NA   |
## | zone3 |  Yes  |  Yes  |  NA   |  NA   |  NA   |  NA   |  NA   |  NA   |
## | zone4 |  Yes  |  No   |  Yes  |  NA   |  NA   |  NA   |  NA   |  NA   |
## | zone5 |  Yes  |  Yes  |  Yes  |  Yes  |  NA   |  NA   |  NA   |  NA   |
## | zone6 |  Yes  |  Yes  |  Yes  |  Yes  |  No   |  NA   |  NA   |  NA   |
## | zone7 |  Yes  |  Yes  |  No   |  Yes  |  Yes  |  Yes  |  NA   |  NA   |
## | zone8 |  Yes  |  Yes  |  No   |  Yes  |  Yes  |  Yes  |  No   |  NA   |
## | zone9 |  Yes  |  Yes  |  No   |  Yes  |  Yes  |  Yes  |  No   |  Yes  |

Ecological detector

ed = gd(incidence ~ watershed + elevation + soiltype,
        data = NTDs,type = 'ecological')
ed
##                 Ecological Detector         
## 
## |          | elevation | soiltype |
## |:---------|:---------:|:--------:|
## |watershed |    No     |    No    |
## |elevation |    NA     |    No    |
plot(ed)

You can also change the significant interval by assign alpha argument,the default value of alpha argument is 0.95.

ed99 = gd(incidence ~ watershed + elevation + soiltype,
          data = NTDs,type = 'ecological',alpha = 0.99)
ed99
##                 Ecological Detector         
## 
## |          | elevation | soiltype |
## |:---------|:---------:|:--------:|
## |watershed |    No     |    No    |
## |elevation |    NA     |    No    |

Running four basic geodetectors simultaneously

g = gd(incidence ~ watershed + elevation + soiltype,
       data = NTDs,
       type = c("factor", "interaction", "risk", "ecological"))
g
##                 Factor Detector            
## 
## | variable  | Q-statistic |   P-value   |
## |:---------:|:-----------:|:-----------:|
## | watershed |  0.6377737  | 0.000128803 |
## | elevation |  0.6067087  | 0.043382244 |
## | soiltype  |  0.3857168  | 0.372145486 |
## 
##                 Interaction Detector         
## 
## | Interactive variable  | Interaction  |
## |:---------------------:|:------------:|
## | watershed ∩ elevation | Enhance, bi- |
## | watershed ∩ soiltype  | Enhance, bi- |
## | elevation ∩ soiltype  | Enhance, bi- |
## 
##                 Risk Detector            
## 
##  Variable elevation:
## 
## | zone  | zone1 | zone2 | zone3 | zone4 | zone5 | zone6 |
## |:-----:|:-----:|:-----:|:-----:|:-----:|:-----:|:-----:|
## | zone2 |  Yes  |  NA   |  NA   |  NA   |  NA   |  NA   |
## | zone3 |  Yes  |  No   |  NA   |  NA   |  NA   |  NA   |
## | zone4 |  Yes  |  Yes  |  Yes  |  NA   |  NA   |  NA   |
## | zone5 |  Yes  |  Yes  |  Yes  |  Yes  |  NA   |  NA   |
## | zone6 |  Yes  |  Yes  |  Yes  |  Yes  |  Yes  |  NA   |
## | zone7 |  Yes  |  Yes  |  Yes  |  Yes  |  Yes  |  Yes  |
## 
##  Variable soiltype:
## 
## | zone  | zone1 | zone2 | zone3 | zone4 |
## |:-----:|:-----:|:-----:|:-----:|:-----:|
## | zone2 |  Yes  |  NA   |  NA   |  NA   |
## | zone3 |  Yes  |  No   |  NA   |  NA   |
## | zone4 |  Yes  |  Yes  |  Yes  |  NA   |
## | zone5 |  No   |  Yes  |  Yes  |  Yes  |
## 
##  Variable watershed:
## 
## | zone  | zone1 | zone2 | zone3 | zone4 | zone5 | zone6 | zone7 | zone8 |
## |:-----:|:-----:|:-----:|:-----:|:-----:|:-----:|:-----:|:-----:|:-----:|
## | zone2 |  Yes  |  NA   |  NA   |  NA   |  NA   |  NA   |  NA   |  NA   |
## | zone3 |  Yes  |  Yes  |  NA   |  NA   |  NA   |  NA   |  NA   |  NA   |
## | zone4 |  Yes  |  No   |  Yes  |  NA   |  NA   |  NA   |  NA   |  NA   |
## | zone5 |  Yes  |  Yes  |  Yes  |  Yes  |  NA   |  NA   |  NA   |  NA   |
## | zone6 |  Yes  |  Yes  |  Yes  |  Yes  |  No   |  NA   |  NA   |  NA   |
## | zone7 |  Yes  |  Yes  |  No   |  Yes  |  Yes  |  Yes  |  NA   |  NA   |
## | zone8 |  Yes  |  Yes  |  No   |  Yes  |  Yes  |  Yes  |  No   |  NA   |
## | zone9 |  Yes  |  Yes  |  No   |  Yes  |  Yes  |  Yes  |  No   |  Yes  |
## 
##                 Ecological Detector         
## 
## |          | elevation | soiltype |
## |:---------|:---------:|:--------:|
## |watershed |    No     |    No    |
## |elevation |    NA     |    No    |
plot(g)

These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.