Regresi Spasial Praktikum 11

A. Model Prediksi GRP Data China

1. Persiapan Data

library(sf)
## Warning: package 'sf' was built under R version 4.5.3
## Linking to GEOS 3.14.1, GDAL 3.12.1, PROJ 9.7.1; sf_use_s2() is TRUE
library(readxl)
## Warning: package 'readxl' was built under R version 4.5.3
library(tidyverse)
## Warning: package 'tidyverse' was built under R version 4.5.3
## Warning: package 'ggplot2' was built under R version 4.5.3
## Warning: package 'tibble' was built under R version 4.5.3
## Warning: package 'tidyr' was built under R version 4.5.3
## Warning: package 'readr' was built under R version 4.5.3
## Warning: package 'purrr' was built under R version 4.5.3
## Warning: package 'dplyr' was built under R version 4.5.3
## Warning: package 'stringr' was built under R version 4.5.3
## Warning: package 'forcats' was built under R version 4.5.3
## Warning: package 'lubridate' was built under R version 4.5.3
## ── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
## ✔ dplyr     1.2.1     ✔ readr     2.2.0
## ✔ forcats   1.0.1     ✔ stringr   1.6.0
## ✔ ggplot2   4.0.3     ✔ tibble    3.3.1
## ✔ lubridate 1.9.5     ✔ tidyr     1.3.2
## ✔ purrr     1.2.2
## ── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
## ✖ dplyr::filter() masks stats::filter()
## ✖ dplyr::lag()    masks stats::lag()
## ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errors
library(ggplot2)
library(spdep)
## Warning: package 'spdep' was built under R version 4.5.3
## Loading required package: spData
## Warning: package 'spData' was built under R version 4.5.3
## To access larger datasets in this package, install the spDataLarge
## package with: `install.packages('spDataLarge',
## repos='https://nowosad.github.io/drat/', type='source')`
library(spatialreg)
## Warning: package 'spatialreg' was built under R version 4.5.3
## Loading required package: Matrix
## 
## Attaching package: 'Matrix'
## 
## The following objects are masked from 'package:tidyr':
## 
##     expand, pack, unpack
## 
## 
## Attaching package: 'spatialreg'
## 
## The following objects are masked from 'package:spdep':
## 
##     get.ClusterOption, get.coresOption, get.mcOption,
##     get.VerboseOption, get.ZeroPolicyOption, set.ClusterOption,
##     set.coresOption, set.mcOption, set.VerboseOption,
##     set.ZeroPolicyOption
library(MuMIn)
## Warning: package 'MuMIn' was built under R version 4.5.3
## Registered S3 method overwritten by 'MuMIn':
##   method     from      
##   nobs.Sarlm spatialreg
library(patchwork)
## Warning: package 'patchwork' was built under R version 4.5.3
# Membaca data spasial dalam format shapefile (.shp)
shp_china <- st_read("C:/Users/Abit/Documents/KULIAH SMT 5/REGSPAS/Regspas_11/Regspas_11/data china/China29/China29.shp")
## Reading layer `China29' from data source 
##   `C:\Users\Abit\Documents\KULIAH SMT 5\REGSPAS\Regspas_11\Regspas_11\data china\China29\China29.shp' 
##   using driver `ESRI Shapefile'
## Simple feature collection with 29 features and 12 fields
## Geometry type: MULTIPOLYGON
## Dimension:     XY
## Bounding box:  xmin: 73.5577 ymin: 20.22264 xmax: 134.7739 ymax: 53.56086
## Geodetic CRS:  WGS 84
# Membaca data Excel
data_china <- read_excel("C:/Users/Abit/Documents/KULIAH SMT 5/REGSPAS/Regspas_11/Regspas_11/data china/Data_China.xlsx")

# Menggabungkan data spasial dengan data Excel
china_join <- shp_china %>%
  left_join(data_china,
            by = c("NAME_1" = "Region"))

# Melihat hasil penggabungan
head(china_join)
## Simple feature collection with 6 features and 16 fields
## Geometry type: MULTIPOLYGON
## Dimension:     XY
## Bounding box:  xmin: 92.76416 ymin: 20.22264 xmax: 120.8376 ymax: 42.77393
## Geodetic CRS:  WGS 84
##   ID_0 ISO NAME_0 ID_1    NAME_1 HASC_1 CCN_1 CCA_1    TYPE_1    ENGTYPE_1
## 1   49 CHN  China    1     Anhui  CN.AH     0  <NA>     Sh?ng     Province
## 2   49 CHN  China    2   Beijing  CN.BJ     0  <NA> Zhíxiáshì Municipality
## 3   49 CHN  China    3 Chongqing  CN.CQ     0  <NA> Zhíxiáshì Municipality
## 4   49 CHN  China    4    Fujian  CN.FJ     0  <NA>     Sh?ng     Province
## 5   49 CHN  China    5     Gansu  CN.GS     0  <NA>     Sh?ng     Province
## 6   49 CHN  China    6 Guangdong  CN.GD     0  <NA>     Sh?ng     Province
##   NL_NAME_1 VARNAME_1     Long      Lat   GRP Level_of_Urbanization
## 1     ??|??     ?nhu? 117.2262 31.82579 28792                 46.50
## 2     ??|??   B?ij?ng 116.4107 40.18491 87475                 86.20
## 3     ??|?? Chóngqìng 107.8748 30.05865 38914                 56.98
## 4        ??    Fújiàn 117.9815 26.08641 52763                 59.60
## 5     ??|??     G?nsù 100.9339 37.82079 21978                 38.75
## 6     ??|?? Gu?ngd?ng 113.4164 23.34225 54095                 67.40
##                         geometry
## 1 MULTIPOLYGON (((116.4263 34...
## 2 MULTIPOLYGON (((116.6669 40...
## 3 MULTIPOLYGON (((108.5419 32...
## 4 MULTIPOLYGON (((117.689 23....
## 5 MULTIPOLYGON (((97.18472 42...
## 6 MULTIPOLYGON (((110.1182 20...

2. Eksplorasi Data

GRP <- ggplot(data = china_join) +
  geom_sf(
    aes(fill = GRP),
    color = "white",
    linewidth = 0.2
  ) +
  
  scale_fill_viridis_c(
    option = "magma",
    name = "GRP"
  ) +
  
  labs(
    title = "Gross Regional Product",
    subtitle = "Data kontinu per provinsi",
    x = "Longitude",
    y = "Latitude"
  ) +
  
  theme_minimal() +
  
  theme(
    legend.position = "bottom",
    plot.title = element_text(face = "bold", size = 14),
    plot.subtitle = element_text(size = 11)
  )

Urbanization <- ggplot(data = china_join) +
  geom_sf(
    aes(fill = `Level_of_Urbanization`),
    color = "white",
    linewidth = 0.2
  ) +
  
  scale_fill_viridis_c(
    option = "magma",
    name = "Level of Urbanization"
  ) +
  
  labs(
    title = "Level of Urbanization",
    subtitle = "Data kontinu per provinsi",
    x = "Longitude",
    y = "Latitude"
  ) +
  
  theme_minimal() +
  
  theme(
    legend.position = "bottom",
    plot.title = element_text(face = "bold", size = 14),
    plot.subtitle = element_text(size = 11)
  )

GRP | Urbanization

## 3. Regresi Spasial

china.lm<- lm(GRP ~ Level_of_Urbanization,data=china_join)
summary(china.lm)
## 
## Call:
## lm(formula = GRP ~ Level_of_Urbanization, data = china_join)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -12050.9  -4410.4   -374.8   2961.6  14948.1 
## 
## Coefficients:
##                       Estimate Std. Error t value Pr(>|t|)    
## (Intercept)           -31895.5     5197.6  -6.137 1.48e-06 ***
## Level_of_Urbanization   1400.0       92.6  15.118 1.06e-14 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 6595 on 27 degrees of freedom
## Multiple R-squared:  0.8944, Adjusted R-squared:  0.8904 
## F-statistic: 228.6 on 1 and 27 DF,  p-value: 1.064e-14

4. Matriks Pembobot Spasial

# Baca jarak kereta dari sheet Railway Distance
rail <- read_excel("C:/Users/Abit/Documents/KULIAH SMT 5/REGSPAS/Regspas_11/Regspas_11/data china/Data_China.xlsx",
                   sheet = "Railway Distance",
                   skip = 1)
## New names:
## • `` -> `...1`
rail_matrix <- as.matrix(rail[, 3:31])
rownames(rail_matrix) <- colnames(rail_matrix)

# Inverse distance (w_ij = 1/d_ij, diagonal = 0)
W_rail <- 1 / rail_matrix
diag(W_rail) <- 0

# Row-standardize
W_rail_std <- W_rail / rowSums(W_rail)

# Buat listw object
W <- mat2listw(W_rail_std, style = "W")

5. Uji Morans’I

Untuk memahami pola hubungan spasial antara tingkat urbanisasi (Level of Urbanization) dan pertumbuhan ekonomi daerah (Gross Regional Product - GRP) di China, dilakukan pengujian autokorelasi spasial pada sisaan (residuals) model OLS dengan menggunakan Uji Moran.

col.moran <- lm.morantest(china.lm, W)
col.moran
## 
##  Global Moran I for regression residuals
## 
## data:  
## model: lm(formula = GRP ~ Level_of_Urbanization, data = china_join)
## weights: W
## 
## Moran I statistic standard deviate = 1.1234, p-value = 0.1306
## alternative hypothesis: greater
## sample estimates:
## Observed Moran I      Expectation         Variance 
##      0.010629595     -0.033552435      0.001546711

Berdasarkan hasil uji Global Moran’s I terhadap sisaan model OLS, diperoleh nilai koefisien yang tidak signifikan (p-value>0.05). Hasil ini mengindikasikan bahwa sisaan dari model regresi menyebar secara acak dan tidak memperlihatkan adanya pola mengelompok (clustering) yang terstruktur secara spasial di sepanjang jaringan kereta api China.

6. Uji Lagrange Multiplier

Pengujian Lagrange Multiplier (LM Test). Baik uji LM-Lag maupun LM-Error menghasilkan nilai yang tidak signifikan (p-value>0.05). Tidak signifikannya uji LM-Lag menandakan tidak adanya efek ketergantungan spasial global pada variabel dependen (Global Spatial Lag Dependent). Artinya, tinggi rendahnya nilai GRP di suatu wilayah tidak didorong oleh efek penularan atau umpan balik sistemik dari nilai GRP wilayah lain di sepanjang koridor kereta api tersebut. Sementara itu, nilai LM-Error yang tidak signifikan mengonfirmasi tidak adanya bias dari variabel luar (omitted variables) yang polanya terstruktur secara spasial pada komponen error.

china.lagrange1 <- lm.RStests(china.lm, W, test=c("LMerr","RLMerr","LMlag","RLMlag","SARMA"))
summary(china.lagrange1)
##  Rao's score (a.k.a Lagrange multiplier) diagnostics for spatial
##  dependence
## data:  
## model: lm(formula = GRP ~ Level_of_Urbanization, data = china_join)
## test weights: W
##  
##          statistic parameter p.value
## RSerr    0.0277967         1  0.8676
## adjRSerr 0.0726855         1  0.7875
## RSlag    0.0035418         1  0.9525
## adjRSlag 0.0484307         1  0.8258
## SARMA    0.0762273         2  0.9626

Oleh karena itu, jalur pemodelan spasial global yang mengasumsikan efek multi-arah tak terbatas (seperti Spatial Autoregressive [SAR], Spatial Error Model [SEM], maupun Spatial Durbin Model [SDM]) diputuskan tidak relevan untuk menjelaskan hubungan Urbanisasi dan GRP dalam penelitian ini.

7. SLX Model

china.SLX <- lmSLX(GRP ~ Level_of_Urbanization, data = china_join, listw =  W, Durbin = TRUE);
summary(china.SLX)
## 
## Call:
## lm(formula = formula(paste("y ~ ", paste(colnames(x)[-1], collapse = "+"))), 
##     data = as.data.frame(x), weights = weights)
## 
## Coefficients:
##                            Estimate    Std. Error  t value     Pr(>|t|)  
## (Intercept)                -2.335e+04   4.130e+04  -5.654e-01   5.766e-01
## Level_of_Urbanization       1.387e+03   1.139e+02   1.217e+01   3.074e-12
## lag.Level_of_Urbanization  -1.418e+02   6.801e+02  -2.086e-01   8.364e-01

Level_of_Urbanization berpengaruh positif dan signifikan secara statistik (p < 0,001). Artinya, setiap kenaikan satu satuan tingkat urbanisasi di suatu provinsi akan meningkatkan variabel dependen sebesar 1.387 satuan secara langsung (direct effect). Ini menunjukkan bahwa urbanisasi di wilayah itu sendiri memiliki peran nyata terhadap variabel yang dimodelkan.

Sebaliknya, lag.Level_of_Urbanization yang merepresentasikan rata-rata tingkat urbanisasi provinsi-provinsi tetangga memiliki koefisien negatif (-141,8) namun tidak signifikan (pvalue= 0,836>0.05). Artinya, tingkat urbanisasi di provinsi-provinsi sekitar tidak terbukti memberikan pengaruh spasial (spill-over) yang berarti terhadap variabel dependen.

Untuk membuktikan apakah penambahan komponen spasial lokal (WX) pada model SLX memberikan peningkatan performa yang bermakna dibandingkan model OLS biasa, dilakukan uji formal Likelihood Ratio (LR) Test

8. LR test : SLX vs OLS Regression

# log likelihood masing-masing model
logLik(china.SLX)
## 'log Lik.' -295.1173 (df=4)
logLik(china.lm)
## 'log Lik.' -295.1415 (df=3)
# nilai LR
LR <- 2 * (as.numeric(logLik(china.SLX)) - 
           as.numeric(logLik(china.lm)))

# derajat bebas
df <- attr(logLik(china.SLX), "df") - 
      attr(logLik(china.lm), "df")

# p-value
pchisq(LR, df = df, lower.tail = FALSE)
## [1] 0.8257453
AICc(china.lm)
## [1] 597.2431
AICc(china.SLX)
## [1] 599.9012

Berdasarkan uji perbandingan tersebut, diperoleh nilai Likelihood Ratio yang tidak signifikan (p-value>0.05) yang disertai dengan nilai informasi kriteria AICc yang tidak lebih kecil dibandingkan OLS regression. Hasil statistik formal ini menyatakan gagal tolak H0​, yang berarti penambahan variabel tingkat urbanisasi tetangga berbasis konektivitas kereta api (WX) tidak memberikan kontribusi signifikan dalam menurunkan AICc. Oleh karena itu, model regresi OLS menjadi model terbaik dalam memodelkan pengaruh Level_of_Urbanization terhadap GRP.

B. Model Dependensi Spasial (Jawa Barat)

1. Pre processing Data

# Membaca data spasial shapefile Jawa Barat
shp_jabar <- st_read("C:/Users/Abit/Documents/KULIAH SMT 5/REGSPAS/Regspas_11/Regspas_11/JABAR/Jabar2/Jabar2.shp")
## Reading layer `Jabar2' from data source 
##   `C:\Users\Abit\Documents\KULIAH SMT 5\REGSPAS\Regspas_11\Regspas_11\JABAR\Jabar2\Jabar2.shp' 
##   using driver `ESRI Shapefile'
## Simple feature collection with 26 features and 7 fields
## Geometry type: MULTIPOLYGON
## Dimension:     XY
## Bounding box:  xmin: 106.3705 ymin: -7.823398 xmax: 108.8338 ymax: -5.91377
## CRS:           NA
# Membaca data Excel
data_jabar <- read_excel("C:/Users/Abit/Documents/KULIAH SMT 5/REGSPAS/Regspas_11/Regspas_11/JABAR/data_jabar.xlsx")

# Mengubah KABKOTNO menjadi character
data_jabar$KABKOTNO <- as.character(data_jabar$KABKOTNO)

# Menggabungkan data shapefile dengan data Excel
jabar <- shp_jabar %>%
  left_join(data_jabar, by = "KABKOTNO") %>%
  na.omit()

# Melihat nama kolom
colnames(jabar)
##  [1] "PROVNO.x"   "KABKOTNO"   "KODE2010.x" "PROVINSI.x" "KABKOT.x"  
##  [6] "SUMBER"     "IDSP2010.x" "PROVNO.y"   "KODE2010.y" "PROVINSI.y"
## [11] "KABKOT.y"   "IDSP2010.y" "Long"       "Lat"        "p.miskin15"
## [16] "p.miskin16" "j.miskin15" "j.miskin16" "AHH2015"    "AHH2016"   
## [21] "EYS2015"    "EYS2016"    "MYS2015"    "MYS2016"    "EXP2015"   
## [26] "EXP2016"    "APM.SD15"   "APM.SMP15"  "APM.SMA15"  "APM.PT15"  
## [31] "APK.SD15"   "APK.SMP15"  "APK.SMA15"  "APK.PT15"   "APS.USIA15"
## [36] "APS.USIA2"  "APS.USIA3"  "APS.USIA4"  "geometry"

2. Eksplorasi

plot(jabar[,c(
  "p.miskin15",
  "AHH2015",
  "EYS2015",
  "MYS2015",
  "EXP2015",
  "APM.SD15",
  "APM.SMP15",
  "APM.SMA15",
  "APM.PT15",
  "APK.SD15",
  "APK.SMP15",
  "APK.SMA15",
  "APK.PT15",
  "APS.USIA15",
  "APS.USIA2",
  "APS.USIA3",
  "APS.USIA4"
)])
## Warning: plotting the first 9 out of 17 attributes; use max.plot = 17 to plot
## all

##. 3. Matriks Pembobot Spasial

# =====================================================
# SPATIAL WEIGHT
# =====================================================
nb_jabar <- suppressWarnings(poly2nb(jabar))

lw_jabar <- nb2listw(
  nb_jabar,
  style = "W",
  zero.policy = TRUE
)

4. Penentuan Variabel (Indeks Moran)

# =========================================================
# ANALISIS MORAN'S I RESIDUAL
# MENCARI KOMBINASI X TERBAIK
# DENGAN P-VALUE TERKECIL
# =========================================================

# =========================================================
# 5. DAFTAR VARIABEL
# =========================================================
variabel_list <- c(
  "p.miskin15",
  "AHH2015",
  "EYS2015",
  "MYS2015",
  "EXP2015",
  "APM.SD15",
  "APM.SMP15",
  "APM.SMA15",
  "APM.PT15",
  "APK.SD15",
  "APK.SMP15",
  "APK.SMA15",
  "APK.PT15",
  "APS.USIA15",
  "APS.USIA2",
  "APS.USIA3",
  "APS.USIA4"
)

# =========================================================
# 6. TENTUKAN VARIABEL Y
# =========================================================
variabel_Y <- "p.miskin15"

# variabel kandidat X
variabel_X <- setdiff(
  variabel_list,
  variabel_Y
)

# =========================================================
# 7. DATAFRAME HASIL
# =========================================================
hasil_model <- data.frame(
  Formula = character(),
  Jumlah_X = numeric(),
  Moran_I = numeric(),
  P_Value = numeric(),
  stringsAsFactors = FALSE
)

# =========================================================
# 8. LOOP KOMBINASI VARIABEL
# =========================================================

maksimum_X <- 6

for (k in 1:maksimum_X) {
  # membuat kombinasi
  kombinasi_x <- combn(
    variabel_X,
    k,
    simplify = FALSE
  )
  
  # loop tiap kombinasi
  for (x_set in kombinasi_x) {
    
    tryCatch({
      
      # =====================================================
      # MEMBUAT FORMULA
      # =====================================================
      formula_text <- paste(
        variabel_Y,
        "~",
        paste(x_set, collapse = " + ")
      )
      
      rumus <- as.formula(formula_text)
      
      # =====================================================
      # MODEL OLS
      # =====================================================
      model_ols <- lm(
        rumus,
        data = jabar
      )
      
      # =====================================================
      # MORAN TEST RESIDUAL
      # =====================================================
      uji_moran <- lm.morantest(
        model_ols,
        lw_jabar,
        zero.policy = TRUE
      )
      
      # =====================================================
      # AMBIL NILAI
      # =====================================================
      moran_i <- as.numeric(
        uji_moran$estimate["Observed Moran I"]
      )
      
      p_val <- uji_moran$p.value
      
      # =====================================================
      # SIMPAN JIKA SIGNIFIKAN
      # =====================================================
      if (p_val < 0.05) {
        
        hasil_model <- rbind(
          hasil_model,
          
          data.frame(
            Formula = formula_text,
            Jumlah_X = k,
            Moran_I = moran_i,
            P_Value = p_val
          )
        )
      }
      
    }, error = function(e) {
      
      # abaikan error
      NULL
      
    })
  }
}

# =========================================================
# 9. URUTKAN HASIL
# =========================================================
hasil_model <- hasil_model %>%
  arrange(P_Value)


# =========================================================
# 11. MODEL TERBAIK
# P-VALUE PALING KECIL
# =========================================================


if (nrow(hasil_model) == 0) {
  
  cat("Tidak ada model signifikan.\n")
  
} else {
  
  # ambil model terbaik
  model_terbaik <- hasil_model %>%
    slice(1)
  
  cat(
    "Formula              : ",
    as.character(model_terbaik$Formula),
    "\n",
    sep = ""
  )
  
  cat(
    "Jumlah Variabel X    : ",
    model_terbaik$Jumlah_X,
    "\n",
    sep = ""
  )
  
  cat(
    "Moran's I            : ",
    round(model_terbaik$Moran_I, 5),
    "\n",
    sep = ""
  )
  
  cat(
    "P-Value              : ",
    format(
      model_terbaik$P_Value,
      scientific = FALSE
    ),
    "\n",
    sep = ""
  )
}
## Formula              : p.miskin15 ~ EYS2015 + APM.SD15 + APK.SD15 + APK.PT15 + APS.USIA15
## Jumlah Variabel X    : 5
## Moran's I            : 0.5012
## P-Value              : 0.005018003

Berdasarkan hasil pencarian kombinasi variabel terbaik, terpilih model dengan formula p.miskin15 ~ EYS2015 + APM.SD15 + APK.SD15 + APK.PT15 + APS.USIA15 yang terdiri dari 5 variabel independen. Model ini menghasilkan nilai Moran’s I sebesar 0,5012 dengan p-value 0,005 — signifikan pada taraf 1% — yang menunjukkan bahwa residual model OLS dengan kombinasi variabel ini mengandung autokorelasi spasial positif yang kuat. Artinya, wilayah-wilayah di Jawa Barat dengan pola kemiskinan yang serupa cenderung berkelompok secara geografis dan belum sepenuhnya dijelaskan oleh kelima variabel tersebut secara biasa. Kombinasi ini dipilih sebagai model terbaik karena menghasilkan p-value Moran’s I terkecil di antara seluruh kombinasi yang diuji, sehingga menjadi justifikasi terkuat untuk melanjutkan analisis menggunakan model regresi spasial yang mampu mengakomodasi struktur dependensi spasial tersebut secara eksplisit.

5. OLS Regression

jabar.lm <- lm(p.miskin15 ~ EYS2015 + APM.SD15 + APK.SD15 + APK.PT15 + APS.USIA15, data=jabar)
summary(jabar.lm)
## 
## Call:
## lm(formula = p.miskin15 ~ EYS2015 + APM.SD15 + APK.SD15 + APK.PT15 + 
##     APS.USIA15, data = jabar)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -6.2541 -1.5129  0.7468  1.7253  2.9766 
## 
## Coefficients:
##              Estimate Std. Error t value Pr(>|t|)  
## (Intercept) -263.5186   236.1804  -1.116   0.2883  
## EYS2015        2.7220     2.1113   1.289   0.2238  
## APM.SD15      -0.5289     0.6939  -0.762   0.4619  
## APK.SD15       0.4526     0.2946   1.536   0.1527  
## APK.PT15      -0.3228     0.1731  -1.865   0.0891 .
## APS.USIA15     2.4849     2.5164   0.987   0.3446  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 3.114 on 11 degrees of freedom
## Multiple R-squared:  0.5749, Adjusted R-squared:  0.3816 
## F-statistic: 2.975 on 5 and 11 DF,  p-value: 0.06127

6. Uji Lagrange Multiplier

jabar.lagrange <- lm.LMtests(jabar.lm, lw_jabar, test=c("LMerr","RLMerr","LMlag","RLMlag","SARMA"))
## Please update scripts to use lm.RStests in place of lm.LMtests
summary(jabar.lagrange)
##  Rao's score (a.k.a Lagrange multiplier) diagnostics for spatial
##  dependence
## data:  
## model: lm(formula = p.miskin15 ~ EYS2015 + APM.SD15 + APK.SD15 +
## APK.PT15 + APS.USIA15, data = jabar)
## test weights: listw
##  
##          statistic parameter  p.value   
## RSerr      7.23955         1 0.007132 **
## adjRSerr   5.48939         1 0.019132 * 
## RSlag      2.01129         1 0.156133   
## adjRSlag   0.26113         1 0.609346   
## SARMA      7.50067         2 0.023510 * 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Hasil uji LM menunjukkan bahwa dependensi spasial dalam data Jawa Barat bersumber dari komponen error, bukan dari lag spasial variabel dependen. Uji RSerr signifikan (p = 0,007) mengindikasikan adanya autokorelasi spasial pada residual model, dan ketika dikontrol terhadap pengaruh lag, adjRSerr tetap signifikan (p = 0,019) — membuktikan bahwa spatial error hadir secara mandiri dalam model.

Sebaliknya, uji RSlag tidak signifikan (p = 0,156) dan adjRSlag pun tidak signifikan (p = 0,609), yang berarti tidak ada bukti bahwa tingkat kemiskinan di suatu wilayah dipengaruhi langsung oleh tingkat kemiskinan wilayah tetangganya.

5. Spatial Error Model (SEM)

jabar.err <- errorsarlm(p.miskin15 ~ EYS2015 + APM.SD15 + APK.SD15 + APK.PT15 + APS.USIA15,lw_jabar,data=jabar, zero.policy = TRUE)
summary(jabar.err)
## 
## Call:errorsarlm(formula = p.miskin15 ~ EYS2015 + APM.SD15 + APK.SD15 + 
##     APK.PT15 + APS.USIA15, data = jabar, listw = lw_jabar, zero.policy = TRUE)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -1.96720 -0.97706  0.17825  0.76731  1.45103 
## 
## Type: error 
## Regions with no neighbours included:
##  18 19 21 25 26 
## Coefficients: (asymptotic standard errors) 
##                Estimate  Std. Error z value  Pr(>|z|)
## (Intercept) -425.458866   84.564213 -5.0312 4.874e-07
## EYS2015        4.311729    1.035451  4.1641 3.126e-05
## APM.SD15      -1.521207    0.226462 -6.7173 1.852e-11
## APK.SD15       0.640932    0.101930  6.2880 3.216e-10
## APK.PT15      -0.358322    0.069408 -5.1626 2.436e-07
## APS.USIA15     4.680477    0.828208  5.6513 1.592e-08
## 
## Lambda: 0.9059, LR test value: 23.088, p-value: 1.5476e-06
## Asymptotic standard error: 0.060606
##     z-value: 14.947, p-value: < 2.22e-16
## Wald statistic: 223.43, p-value: < 2.22e-16
## 
## Log likelihood: -28.18863 for error model
## ML residual variance (sigma squared): 1.0473, (sigma: 1.0234)
## Number of observations: 17 
## Number of parameters estimated: 8 
## AIC: 72.377, (AIC for lm: 93.465)

6. Model SLX

jabar.SLX <- lmSLX(p.miskin15 ~ EYS2015 + APM.SD15 + APK.SD15 + APK.PT15 + APS.USIA15,lw_jabar,data=jabar, zero.policy = TRUE, Durbin = TRUE);
summary(jabar.SLX)
## 
## Call:
## lm(formula = formula(paste("y ~ ", paste(colnames(x)[-1], collapse = "+"))), 
##     data = as.data.frame(x), weights = weights)
## 
## Coefficients:
##                 Estimate    Std. Error  t value     Pr(>|t|)  
## (Intercept)     -145.59982   307.35728    -0.47372     0.65245
## EYS2015            2.63817     4.59362     0.57431     0.58663
## APM.SD15           0.26488     0.97666     0.27122     0.79532
## APK.SD15           0.59059     0.23882     2.47293     0.04827
## APK.PT15          -0.20577     0.38277    -0.53759     0.61021
## APS.USIA15         0.36388     3.17195     0.11472     0.91241
## lag.EYS2015       -4.02200     9.31944    -0.43157     0.68113
## lag.APM.SD15       3.42980     1.21392     2.82538     0.03014
## lag.APK.SD15      -0.41090     0.69159    -0.59414     0.57412
## lag.APK.PT15       0.38003     0.82137     0.46268     0.65990
## lag.APS.USIA15    -2.48727     1.03245    -2.40910     0.05264

7. Model SARMA

jabar.sarma <- sacsarlm(p.miskin15 ~ EYS2015 + APM.SD15 + APK.SD15 + APK.PT15 + APS.USIA15,lw_jabar, data=jabar, zero.policy = TRUE)
summary(jabar.sarma)
## 
## Call:sacsarlm(formula = p.miskin15 ~ EYS2015 + APM.SD15 + APK.SD15 + 
##     APK.PT15 + APS.USIA15, data = jabar, listw = lw_jabar, zero.policy = TRUE)
## 
## Residuals:
##       Min        1Q    Median        3Q       Max 
## -1.837342 -0.900247 -0.050073  0.685523  1.606525 
## 
## Type: sac 
## Coefficients: (asymptotic standard errors) 
##                Estimate  Std. Error z value  Pr(>|z|)
## (Intercept) -442.074286   83.148799 -5.3167 1.057e-07
## EYS2015        4.793132    1.140231  4.2037 2.626e-05
## APM.SD15      -1.707710    0.293038 -5.8276 5.623e-09
## APK.SD15       0.672837    0.103147  6.5231 6.886e-11
## APK.PT15      -0.392055    0.076727 -5.1097 3.226e-07
## APS.USIA15     4.939024    0.838588  5.8897 3.869e-09
## 
## Rho: 0.20377
## Asymptotic standard error: 0.19972
##     z-value: 1.0202, p-value: 0.30761
## Lambda: 0.91256
## Asymptotic standard error: 0.065071
##     z-value: 14.024, p-value: < 2.22e-16
## 
## LR test value: 23.918, p-value: 6.4015e-06
## 
## Log likelihood: -27.77358 for sac model
## ML residual variance (sigma squared): 0.97176, (sigma: 0.98578)
## Number of observations: 17 
## Number of parameters estimated: 9 
## AIC: 73.547, (AIC for lm: 93.465)

Model SARMA (tipe sac) menghasilkan temuan yang menarik pada kedua parameter spasialnya. Nilai Rho (ρ) sebesar 0,204 tidak signifikan (p = 0,308), yang berarti tidak terbukti adanya pengaruh langsung dari tingkat kemiskinan wilayah tetangga terhadap kemiskinan suatu wilayah. Sebaliknya, Lambda (λ) sebesar 0,913 sangat signifikan (p < 2,22e-16), mengindikasikan adanya autokorelasi spasial yang sangat kuat pada komponen error. Faktor-faktor yang tidak teramati dalam model cenderung berkorelasi kuat antar wilayah yang berdekatan.

Pola ini ρ tidak signifikan, λ sangat signifikan sehingga mengonfirmasi bahwa SEM lebih tepat dibandingkan SARMA untuk data in. Hal ini juga akan dibuktikan dengan melakukan Uji Likelihood Ratio Test antara Model SEM dan Model SARMA

8. LR Test : SEM vs SARMA

anova(jabar.err,jabar.sarma)
##             Model df    AIC  logLik Test L.Ratio p-value
## jabar.err       1  8 72.377 -28.189    1                
## jabar.sarma     2  9 73.547 -27.774    2 0.83008 0.36225

9. Akaike Information Criterion Corrected (AICc)

AICc(jabar.lm)
## [1] 105.9096
AICc(jabar.err)
## [1] 90.37725
AICc(jabar.sarma)
## [1] 99.26145