Robust Geographical Detector & Robust Interaction Detector(RGD & RID)

Set up python dependence

Install miniconda
Open the conda terminal and perform steps 3 to 6
Create a new conda env gdverse

conda create -n gdverse python=3.9 -y
Activate this env

conda activate gdverse
Install mamba

conda install -c conda-forge mamba -y
Set up python packages

mamba install -c conda-forge numpy==1.23.5 joblib pandas ruptures -y
Configure GDVERSE_PYTHON environment variable
- Temporarily setting up gdverse python dependence environment using Sys.setenv() in the R console or R code like Sys.setenv(GDVERSE_PYTHON = "path/to/python/environment").
- Or using usethis::edit_r_environ() to set long-available gdverse python dependencies environment path.

Load data and package

library(tidyverse)
library(gdverse)

data("ndvi")

names(ndvi)
## [1] "NDVIchange"    "Climatezone"   "Mining"        "Tempchange"    "Precipitation"
## [6] "GDP"           "Popdensity"

Determine discretization use offline change point detection

Climatezone and Mining is the discrete category variables in the ndvi data, we need to discretize others.You can use robust_disc() to discretize them based on offline change point detection.When executing robust_disc(), note that each discvar can use a discnum parameter of its own. You can pass in a vector of equal length to discvar to use different discnum for different discvar.

tictoc::tic()
new.ndvi = robust_disc(NDVIchange ~ .,
                       data = select(ndvi,-c(Climatezone,Mining)),
                       discnum = 10,cores = 6)
tictoc::toc()
## 20.31 sec elapsed

new.ndvi
## # A tibble: 713 × 4
##    Tempchange Precipitation GDP     Popdensity
##    <chr>      <chr>         <chr>   <chr>     
##  1 group1     group5        group7  group6    
##  2 group1     group5        group7  group4    
##  3 group1     group9        group8  group10   
##  4 group1     group5        group1  group3    
##  5 group1     group3        group1  group3    
##  6 group1     group3        group1  group4    
##  7 group1     group5        group7  group6    
##  8 group1     group5        group8  group10   
##  9 group1     group7        group10 group10   
## 10 group1     group7        group9  group10   
## # ℹ 703 more rows

The new.ndvi is the discrete result,we can combine it with NDVIchange,Climatezone and Mining cols in ndvi tibble now.

new.ndvi = bind_cols(select(ndvi,c(NDVIchange,Climatezone,Mining)),new.ndvi)
new.ndvi
## # A tibble: 713 × 7
##    NDVIchange Climatezone Mining Tempchange Precipitation GDP     Popdensity
##         <dbl> <chr>       <fct>  <chr>      <chr>         <chr>   <chr>     
##  1    0.116   Bwk         low    group1     group5        group7  group6    
##  2    0.0178  Bwk         low    group1     group5        group7  group4    
##  3    0.138   Bsk         low    group1     group9        group8  group10   
##  4    0.00439 Bwk         low    group1     group5        group1  group3    
##  5    0.00316 Bwk         low    group1     group3        group1  group3    
##  6    0.00838 Bwk         low    group1     group3        group1  group4    
##  7    0.0335  Bwk         low    group1     group5        group7  group6    
##  8    0.0387  Bwk         low    group1     group5        group8  group10   
##  9    0.0882  Bsk         low    group1     group7        group10 group10   
## 10    0.0690  Bsk         low    group1     group7        group9  group10   
## # ℹ 703 more rows

Run geodetector

Then, we can run geodetector model by gd() function.

g1 = gd(NDVIchange ~ .,data = new.ndvi, type = 'factor')
g1
## ***          Factor Detector            
## 
## |   variable    | Q-statistic | P-value  |
## |:-------------:|:-----------:|:--------:|
## | Precipitation |  0.8933049  | 3.25e-10 |
## |  Climatezone  |  0.8218335  | 7.34e-10 |
## |  Tempchange   |  0.3874615  | 7.59e-10 |
## |  Popdensity   |  0.2622265  | 3.65e-10 |
## |      GDP      |  0.1491878  | 3.63e-10 |
## |    Mining     |  0.1411154  | 6.73e-10 |

g2 = gd(NDVIchange ~ .,data = new.ndvi, type = 'interaction')
g2
## ***        Interaction Detector         
## 
## |    Interactive variable     | Interaction  |
## |:---------------------------:|:------------:|
## |    Climatezone ∩ Mining     | Enhance, bi- |
## |  Climatezone ∩ Tempchange   | Enhance, bi- |
## | Climatezone ∩ Precipitation | Enhance, bi- |
## |      Climatezone ∩ GDP      | Enhance, bi- |
## |  Climatezone ∩ Popdensity   | Enhance, bi- |
## |     Mining ∩ Tempchange     | Enhance, bi- |
## |   Mining ∩ Precipitation    | Enhance, bi- |
## |        Mining ∩ GDP         | Enhance, bi- |
## |     Mining ∩ Popdensity     | Enhance, bi- |
## | Tempchange ∩ Precipitation  | Enhance, bi- |
## |      Tempchange ∩ GDP       | Enhance, bi- |
## |   Tempchange ∩ Popdensity   | Enhance, bi- |
## |     Precipitation ∩ GDP     | Enhance, bi- |
## | Precipitation ∩ Popdensity  | Enhance, bi- |
## |      GDP ∩ Popdensity       | Enhance, bi- |

You can also use `rgd()` in one time to get result above.

ndvi_rgd = rgd(NDVIchange ~ ., data = ndvi,
               discvar = names(select(ndvi,-c(NDVIchange,Climatezone,Mining))),
               discnum = 10, cores = 6, type = c('factor','interaction'))
ndvi_rgd
##                  RGD Model                  
## ***          Factor Detector            
## 
## |   variable    | Q-statistic | P-value  |
## |:-------------:|:-----------:|:--------:|
## | Precipitation |  0.8933049  | 3.25e-10 |
## |  Climatezone  |  0.8218335  | 7.34e-10 |
## |  Tempchange   |  0.3874615  | 7.59e-10 |
## |  Popdensity   |  0.2622265  | 3.65e-10 |
## |      GDP      |  0.1491878  | 3.63e-10 |
## |    Mining     |  0.1411154  | 6.73e-10 |
## 
## ***        Interaction Detector         
## 
## |    Interactive variable     | Interaction  |
## |:---------------------------:|:------------:|
## |    Climatezone ∩ Mining     | Enhance, bi- |
## |  Climatezone ∩ Tempchange   | Enhance, bi- |
## | Climatezone ∩ Precipitation | Enhance, bi- |
## |      Climatezone ∩ GDP      | Enhance, bi- |
## |  Climatezone ∩ Popdensity   | Enhance, bi- |
## |     Mining ∩ Tempchange     | Enhance, bi- |
## |   Mining ∩ Precipitation    | Enhance, bi- |
## |        Mining ∩ GDP         | Enhance, bi- |
## |     Mining ∩ Popdensity     | Enhance, bi- |
## | Tempchange ∩ Precipitation  | Enhance, bi- |
## |      Tempchange ∩ GDP       | Enhance, bi- |
## |   Tempchange ∩ Popdensity   | Enhance, bi- |
## |     Precipitation ∩ GDP     | Enhance, bi- |
## | Precipitation ∩ Popdensity  | Enhance, bi- |
## |      GDP ∩ Popdensity       | Enhance, bi- |

plot(ndvi_rgd)

Run RID

RID allows you to obtain the interactions of all possible combinations of variables. By default, Spatial Fuzzy And Join are used to generate a new spatial partition of the interaction variables.

ndvi_rid = rid(NDVIchange ~ ., data = ndvi,
               discvar = names(select(ndvi,-c(NDVIchange,Climatezone,Mining))),
               discnum = 10, cores = 6)
ndvi_rid
## ***          Robust Interaction Detector       
## 
## |                               varibale                               |    PD     |
## |:--------------------------------------------------------------------:|:---------:|
## |                     Climatezone ∩ Precipitation                      | 0.8983194 |
## |                            Precipitation                             | 0.8933049 |
## |                  Climatezone ∩ Precipitation ∩ GDP                   | 0.8470890 |
## |                 Climatezone ∩ Mining ∩ Precipitation                 | 0.8448562 |
## |               Climatezone ∩ Precipitation ∩ Popdensity               | 0.8426167 |
## |            Climatezone ∩ Precipitation ∩ GDP ∩ Popdensity            | 0.8351028 |
## |               Climatezone ∩ Tempchange ∩ Precipitation               | 0.8343010 |
## |              Climatezone ∩ Mining ∩ Precipitation ∩ GDP              | 0.8302478 |
## |                         Precipitation ∩ GDP                          | 0.8291002 |
## |          Climatezone ∩ Mining ∩ Precipitation ∩ Popdensity           | 0.8266936 |
## |                      Tempchange ∩ Precipitation                      | 0.8262867 |
## |                             Climatezone                              | 0.8218335 |
## |       Climatezone ∩ Mining ∩ Precipitation ∩ GDP ∩ Popdensity        | 0.8206854 |
## |            Climatezone ∩ Tempchange ∩ Precipitation ∩ GDP            | 0.8033886 |
## |                   Precipitation ∩ GDP ∩ Popdensity                   | 0.7984676 |
## |        Climatezone ∩ Tempchange ∩ Precipitation ∩ Popdensity         | 0.7941671 |
## |                         Climatezone ∩ Mining                         | 0.7930330 |
## |                      Precipitation ∩ Popdensity                      | 0.7928397 |
## |          Climatezone ∩ Mining ∩ Tempchange ∩ Precipitation           | 0.7926405 |
## |     Climatezone ∩ Tempchange ∩ Precipitation ∩ GDP ∩ Popdensity      | 0.7914425 |
## |                   Tempchange ∩ Precipitation ∩ GDP                   | 0.7898127 |
## |       Climatezone ∩ Mining ∩ Tempchange ∩ Precipitation ∩ GDP        | 0.7826698 |
## |                     Mining ∩ Precipitation ∩ GDP                     | 0.7780201 |
## |    Climatezone ∩ Mining ∩ Tempchange ∩ Precipitation ∩ Popdensity    | 0.7767463 |
## | Climatezone ∩ Mining ∩ Tempchange ∩ Precipitation ∩ GDP ∩ Popdensity | 0.7748192 |
## |                        Mining ∩ Precipitation                        | 0.7742147 |
## |                       Climatezone ∩ Tempchange                       | 0.7593504 |
## |            Tempchange ∩ Precipitation ∩ GDP ∩ Popdensity             | 0.7592357 |
## |               Tempchange ∩ Precipitation ∩ Popdensity                | 0.7564501 |
## |                  Climatezone ∩ Mining ∩ Popdensity                   | 0.7554556 |
## |                      Climatezone ∩ Mining ∩ GDP                      | 0.7546429 |
## |              Mining ∩ Precipitation ∩ GDP ∩ Popdensity               | 0.7529482 |
## |                  Climatezone ∩ Mining ∩ Tempchange                   | 0.7485414 |
## |                       Climatezone ∩ Popdensity                       | 0.7478818 |
## |                          Climatezone ∩ GDP                           | 0.7442993 |
## |                 Mining ∩ Precipitation ∩ Popdensity                  | 0.7428155 |
## |              Mining ∩ Tempchange ∩ Precipitation ∩ GDP               | 0.7390511 |
## |                    Climatezone ∩ Tempchange ∩ GDP                    | 0.7373548 |
## |                 Mining ∩ Tempchange ∩ Precipitation                  | 0.7348493 |
## |               Climatezone ∩ Mining ∩ GDP ∩ Popdensity                | 0.7337329 |
## |               Climatezone ∩ Mining ∩ Tempchange ∩ GDP                | 0.7316833 |
## |                    Climatezone ∩ GDP ∩ Popdensity                    | 0.7217405 |
## |                Climatezone ∩ Tempchange ∩ Popdensity                 | 0.7203489 |
## |            Climatezone ∩ Mining ∩ Tempchange ∩ Popdensity            | 0.7190660 |
## |        Mining ∩ Tempchange ∩ Precipitation ∩ GDP ∩ Popdensity        | 0.7124629 |
## |         Climatezone ∩ Mining ∩ Tempchange ∩ GDP ∩ Popdensity         | 0.7106643 |
## |             Climatezone ∩ Tempchange ∩ GDP ∩ Popdensity              | 0.7100158 |
## |           Mining ∩ Tempchange ∩ Precipitation ∩ Popdensity           | 0.7016658 |
## |                Mining ∩ Tempchange ∩ GDP ∩ Popdensity                | 0.4132554 |
## |                      Mining ∩ Tempchange ∩ GDP                       | 0.4057062 |
## |                         Mining ∩ Tempchange                          | 0.4032055 |
## |                   Mining ∩ Tempchange ∩ Popdensity                   | 0.4026808 |
## |                    Tempchange ∩ GDP ∩ Popdensity                     | 0.3996984 |
## |                       Tempchange ∩ Popdensity                        | 0.3990126 |
## |                           Tempchange ∩ GDP                           | 0.3914879 |
## |                              Tempchange                              | 0.3874615 |
## |                      Mining ∩ GDP ∩ Popdensity                       | 0.3293260 |
## |                         Mining ∩ Popdensity                          | 0.3188582 |
## |                           GDP ∩ Popdensity                           | 0.2773077 |
## |                              Popdensity                              | 0.2622265 |
## |                             Mining ∩ GDP                             | 0.2157334 |
## |                                 GDP                                  | 0.1491878 |
## |                                Mining                                | 0.1411154 |