title: “Analisis Regresi Data Panel Statis: Data WDI 15 Negara”
author: “Afrodita E. Pendong”
date: “2026-11-19”
output:
html_document:
toc: true
toc_depth: 3
toc_float:
collapsed: false
smooth_scroll: true
theme: flatly
highlight: tango
number_sections: true
code_folding: show
Persiapan Lingkungan
dan Import Data
Pada tahap ini dilakukan pemanggilan paket (library) yang
dibutuhkan, pembacaan file Excel WDI, serta transformasi struktur data
dari format melebar (wide) menjadi format panel
(long).
library(plm) # Regresi data panel, uji CD, CIPS, Hausman, Chow, LM
library(lmtest) # Uji heteroskedastisitas & autokorelasi
library(sandwich) # Estimasi Robust Standard Error (Driscoll-Kraay / Arellano)
library(car) # Uji multikolinearitas (VIF)
library(tseries) # Uji normalitas Jarque-Bera
library(readxl) # Membaca file Excel (.xlsx)
library(tidyr) # Mengubah bentuk data (Wide ke Long Panel Format)
library(dplyr) # Manipulasi dan pembersihan data
library(knitr) # Menyajikan tabel output HTML yang rapi
file_path <- "C:/Users/User/Downloads/Data WDI 15 Negara.xlsx"
# Membaca Excel (na = ".." otomatis mengubah simbol kosong WDI menjadi NA)
data_raw <- read_excel(file_path, sheet = 1, na = "..")
# Pembersihan dan konversi Wide Format ke Long Panel Format
data_panel <- data_raw %>%
filter(!is.na(Negara), !is.na(Indikator)) %>%
mutate(across(`2000`:`2024`, ~ as.numeric(as.character(.)))) %>%
pivot_longer(
cols = `2000`:`2024`,
names_to = "Tahun",
values_to = "Nilai"
) %>%
mutate(Tahun = as.numeric(Tahun)) %>%
pivot_wider(
names_from = Indikator,
values_from = Nilai,
values_fn = ~ mean(.x, na.rm = TRUE)
) %>%
as.data.frame()
# Penamaan variabel bersih (Y, X1, X2, ...) & pembuatan tabel kamus
indikator_asli <- setdiff(names(data_panel), c("Negara", "Tahun"))
urutan_Y <- 1 # Ubah angka ini jika variabel Y bukan indikator pertama
nama_Y_asli <- indikator_asli[urutan_Y]
nama_X_asli <- indikator_asli[-urutan_Y]
kode_X <- paste0("X", seq_along(nama_X_asli))
kode_semua <- c("Y", kode_X)
kamus_variabel <- data.frame(
Kode = kode_semua,
Indikator_WDI = c(nama_Y_asli, nama_X_asli),
Peran = c("Variabel Dependen (Y)", rep("Variabel Independen (X)", length(kode_X)))
)
kable(kamus_variabel, caption = "Tabel 1. Kamus Kode Variabel Penelitian")
Tabel 1. Kamus Kode Variabel Penelitian
| Y |
Gross fixed capital formation (% of GDP) |
Variabel Dependen (Y) |
| X1 |
Employment in agriculture (% of total employment)
(modeled ILO estimate) |
Variabel Independen (X) |
| X2 |
Deposit interest rate (%) |
Variabel Independen (X) |
| X3 |
Total natural resources rents (% of GDP) |
Variabel Independen (X) |
Pemeriksaan
Kelengkapan Observasi per Negara
data_bersih <- data_panel[, c("Negara", "Tahun", nama_Y_asli, nama_X_asli)]
names(data_bersih) <- c("Negara", "Tahun", kode_semua)
ringkasan_negara <- data_bersih %>%
group_by(Negara) %>%
summarise(
Total_Tahun = n(),
Observasi_Lengkap = sum(complete.cases(across(all_of(kode_semua)))),
Status = ifelse(Observasi_Lengkap > 0, "Masuk Estimasi", "Tereliminasi (Data Kosong)")
)
kable(ringkasan_negara, caption = "Tabel 2. Kelengkapan Data Observasi per Negara (2000-2024)")
Tabel 2. Kelengkapan Data Observasi per Negara
(2000-2024)
| China |
25 |
22 |
Masuk Estimasi |
| India |
25 |
0 |
Tereliminasi (Data Kosong) |
| Indonesia |
25 |
22 |
Masuk Estimasi |
| Japan |
25 |
18 |
Masuk Estimasi |
| Korea, Rep. |
25 |
22 |
Masuk Estimasi |
| Malaysia |
25 |
22 |
Masuk Estimasi |
| Mongolia |
25 |
22 |
Masuk Estimasi |
| Nepal |
25 |
0 |
Tereliminasi (Data Kosong) |
| Philippines |
25 |
20 |
Masuk Estimasi |
| Qatar |
25 |
0 |
Tereliminasi (Data Kosong) |
| Saudi Arabia |
25 |
0 |
Tereliminasi (Data Kosong) |
| Singapore |
25 |
22 |
Masuk Estimasi |
| Thailand |
25 |
20 |
Masuk Estimasi |
| Turkiye |
25 |
22 |
Masuk Estimasi |
| Viet Nam |
25 |
22 |
Masuk Estimasi |
# Deklarasi struktur panel dan formula regresi
pdata <- pdata.frame(data_bersih, index = c("Negara", "Tahun"))
pdim(pdata)
Balanced Panel: n = 15, T = 25, N = 375
model_formula <- as.formula(paste0("Y ~ ", paste0(kode_X, collapse = " + ")))
Statistik
Deskriptif
sampel_estimasi <- data_bersih %>% filter(complete.cases(across(all_of(kode_semua))))
stat_desc <- data.frame(
Variabel = kode_semua,
N = sapply(sampel_estimasi[kode_semua], length),
Mean = round(sapply(sampel_estimasi[kode_semua], mean), 4),
Std_Dev = round(sapply(sampel_estimasi[kode_semua], sd), 4),
Min = round(sapply(sampel_estimasi[kode_semua], min), 4),
Median = round(sapply(sampel_estimasi[kode_semua], median), 4),
Max = round(sapply(sampel_estimasi[kode_semua], max), 4)
)
kable(stat_desc, row.names = FALSE, caption = "Tabel 3. Statistik Deskriptif Variabel Penelitian")
Tabel 3. Statistik Deskriptif Variabel Penelitian
| Y |
234 |
27.9582 |
6.1125 |
18.0765 |
27.2528 |
48.4123 |
| X1 |
234 |
25.5568 |
16.8891 |
0.1078 |
29.0540 |
64.2068 |
| X2 |
234 |
6.3260 |
8.4284 |
0.0356 |
3.3271 |
74.6992 |
| X3 |
234 |
4.6509 |
6.6936 |
0.0002 |
1.9488 |
42.2167 |
Uji Ketergantungan
Lintas Negara (CD) dan Heterogenitas Slope
Uji Cross-Sectional
Dependence (Pesaran CD & Breusch-Pagan LM)
cd_pesaran <- pcdtest(model_formula, data = pdata, test = "cd")
cd_lm <- pcdtest(model_formula, data = pdata, test = "lm")
tabel_cd <- data.frame(
Metode_Uji = c("Pesaran CD Test", "Breusch-Pagan LM Test"),
Statistik = round(c(cd_pesaran$statistic, cd_lm$statistic), 4),
P_Value = format.pval(c(cd_pesaran$p.value, cd_lm$p.value), digits = 4),
Kesimpulan = ifelse(
c(cd_pesaran$p.value, cd_lm$p.value) < 0.05,
"Terdapat Cross-Sectional Dependence (Signifikan)",
"Tidak Ada Cross-Sectional Dependence (Independen)"
)
)
kable(tabel_cd, row.names = FALSE, caption = "Tabel 4. Hasil Uji Cross-Sectional Dependence")
Tabel 4. Hasil Uji Cross-Sectional Dependence
| Pesaran CD Test |
-1.3254 |
0.185 |
Tidak Ada Cross-Sectional Dependence (Independen) |
| Breusch-Pagan LM Test |
106.6902 |
3.678e-05 |
Terdapat Cross-Sectional Dependence (Signifikan) |
Uji Heterogenitas
Slope (Poolability & Swamy Test)
uji_pool <- pooltest(model_formula, data = pdata, model = "within")
pvcm_model <- pvcm(model_formula, data = pdata, model = "random")
sum_pvcm <- summary(pvcm_model)
print(uji_pool)
F statistic
data: model_formula
F = 6.1119, df1 = 30, df2 = 190, p-value = 1.206e-15
alternative hypothesis: unstability
Oneway (individual) effect Random coefficients model
Call:
pvcm(formula = model_formula, data = pdata, model = "random")
Unbalanced Panel: n = 11, T = 18-22, N = 234
Residuals:
Min. 1st Qu. Median Mean 3rd Qu. Max.
-354.417 -61.581 -20.711 -43.965 -6.187 5.226
Estimated mean of the coefficients:
Estimate Std. Error z-value Pr(>|z|)
(Intercept) 30.93443 4.77704 6.4756 9.441e-11 ***
X1 -0.10983 0.49194 -0.2233 0.8233
X2 0.45388 0.64338 0.7055 0.4805
X3 8.79926 16.30122 0.5398 0.5893
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Estimated variance of the coefficients:
(Intercept) X1 X2 X3
(Intercept) 236.7007 -7.7732 -3.0688 -344.358
X1 -7.7732 2.5116 -2.5133 42.910
X2 -3.0688 -2.5133 4.1876 -42.603
X3 -344.3577 42.9097 -42.6027 2745.063
Total Sum of Squares: 209020
Residual Sum of Squares: 1236100
Multiple R-Squared: -4.914
Chisq: 1.27878 on 3 DF, p-value: 0.73418
Test for parameter homogeneity: Chisq = 3477.58 on 40 DF, p-value: < 2.22e-16
Uji Stasioneritas Panel
Generasi Kedua (Pesaran CIPS)
hasil_cips <- lapply(kode_semua, function(v) {
cips_lvl <- cipstest(pdata[[v]], lags = 2, type = "drift")
cips_diff <- cipstest(diff(pdata[[v]]), lags = 2, type = "drift")
data.frame(
Variabel = v,
Stat_Level = round(cips_lvl$statistic, 4),
PVal_Level = round(cips_lvl$p.value, 4),
Status_Level = ifelse(cips_lvl$p.value < 0.05, "Stasioner I(0)", "Tidak Stasioner"),
Stat_Diff1 = round(cips_diff$statistic, 4),
PVal_Diff1 = round(cips_diff$p.value, 4),
Status_Diff1 = ifelse(cips_diff$p.value < 0.05, "Stasioner I(1)", "Tidak Stasioner")
)
}) %>% bind_rows()
kable(hasil_cips, caption = "Tabel 5. Rangkuman Uji Stasioneritas Panel Pesaran CIPS")
Tabel 5. Rangkuman Uji Stasioneritas Panel Pesaran
CIPS
| CIPS test…1 |
Y |
-1.9623 |
0.1 |
Tidak Stasioner |
-2.0521 |
0.1000 |
Tidak Stasioner |
| CIPS test…2 |
X1 |
-1.5972 |
0.1 |
Tidak Stasioner |
-1.9325 |
0.1000 |
Tidak Stasioner |
| CIPS test…3 |
X2 |
-1.3232 |
0.1 |
Tidak Stasioner |
-2.5033 |
0.0205 |
Stasioner I(1) |
| CIPS test…4 |
X3 |
-1.1631 |
0.1 |
Tidak Stasioner |
-2.3437 |
0.0370 |
Stasioner I(1) |
Estimasi Model Regresi
Panel Statis (CEM, FEM, REM)
cem <- plm(model_formula, data = pdata, model = "pooling")
fem <- plm(model_formula, data = pdata, model = "within")
rem <- plm(model_formula, data = pdata, model = "random")
cat("=== 1. COMMON EFFECT MODEL (CEM) ===\n")
=== 1. COMMON EFFECT MODEL (CEM) ===
Pooling Model
Call:
plm(formula = model_formula, data = pdata, model = "pooling")
Unbalanced Panel: n = 11, T = 18-22, N = 234
Residuals:
Min. 1st Qu. Median 3rd Qu. Max.
-10.005 -4.121 -0.459 3.614 15.557
Coefficients:
Estimate Std. Error t-value Pr(>|t|)
(Intercept) 26.735696 0.713379 37.4775 < 2.2e-16 ***
X1 0.057158 0.026395 2.1655 0.031381 *
X2 -0.133543 0.048303 -2.7647 0.006159 **
X3 0.130412 0.064050 2.0361 0.042886 *
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Total Sum of Squares: 8705.5
Residual Sum of Squares: 8064
R-Squared: 0.073688
Adj. R-Squared: 0.061606
F-statistic: 6.09881 on 3 and 230 DF, p-value: 0.00052105
cat("\n=== 2. FIXED EFFECT MODEL (FEM) ===\n")
=== 2. FIXED EFFECT MODEL (FEM) ===
Oneway (individual) effect Within Model
Call:
plm(formula = model_formula, data = pdata, model = "within")
Unbalanced Panel: n = 11, T = 18-22, N = 234
Residuals:
Min. 1st Qu. Median 3rd Qu. Max.
-10.8863 -1.5839 -0.0682 1.6895 13.5473
Coefficients:
Estimate Std. Error t-value Pr(>|t|)
X1 -0.191478 0.039391 -4.8609 2.222e-06 ***
X2 -0.143087 0.046021 -3.1092 0.002124 **
X3 0.362574 0.055984 6.4763 6.054e-10 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Total Sum of Squares: 3000.7
Residual Sum of Squares: 2140.8
R-Squared: 0.28656
Adj. R-Squared: 0.2444
F-statistic: 29.4553 on 3 and 220 DF, p-value: 4.7556e-16
cat("\n=== 3. RANDOM EFFECT MODEL (REM) ===\n")
=== 3. RANDOM EFFECT MODEL (REM) ===
Oneway (individual) effect Random Effect Model
(Swamy-Arora's transformation)
Call:
plm(formula = model_formula, data = pdata, model = "random")
Unbalanced Panel: n = 11, T = 18-22, N = 234
Effects:
var std.dev share
idiosyncratic 9.731 3.119 0.222
individual 34.113 5.841 0.778
theta:
Min. 1st Qu. Median Mean 3rd Qu. Max.
0.8751 0.8869 0.8869 0.8850 0.8869 0.8869
Residuals:
Min. 1st Qu. Median Mean 3rd Qu. Max.
-10.57539 -1.73041 -0.01361 0.00744 1.79855 13.65333
Coefficients:
Estimate Std. Error z-value Pr(>|z|)
(Intercept) 31.375516 1.989225 15.7727 < 2.2e-16 ***
X1 -0.167271 0.037607 -4.4479 8.671e-06 ***
X2 -0.148625 0.045451 -3.2700 0.001076 **
X3 0.359311 0.055549 6.4684 9.905e-11 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Total Sum of Squares: 3067.7
Residual Sum of Squares: 2258
R-Squared: 0.26398
Adj. R-Squared: 0.25438
Chisq: 83.2666 on 3 DF, p-value: < 2.22e-16
Pemilihan Model Terbaik
(Chow, Hausman, Lagrange Multiplier)
chow_test <- pFtest(fem, cem)
hausman_test <- phtest(fem, rem)
lm_test <- plmtest(cem, type = "bp")
if (chow_test$p.value < 0.05) {
if (hausman_test$p.value < 0.05) {
nama_terpilih <- "Fixed Effect Model (FEM)"
model_terpilih <- fem
} else {
nama_terpilih <- "Random Effect Model (REM)"
model_terpilih <- rem
}
} else {
if (lm_test$p.value < 0.05) {
nama_terpilih <- "Random Effect Model (REM)"
model_terpilih <- rem
} else {
nama_terpilih <- "Common Effect Model (CEM)"
model_terpilih <- cem
}
}
tabel_pemilihan <- data.frame(
Tahap_Uji = c("1. Uji Chow (CEM vs FEM)", "2. Uji Hausman (REM vs FEM)", "3. Uji Breusch-Pagan LM (CEM vs REM)"),
Statistik = round(c(chow_test$statistic, hausman_test$statistic, lm_test$statistic), 4),
P_Value = format.pval(c(chow_test$p.value, hausman_test$p.value, lm_test$p.value), digits = 4),
Rekomendasi = c(
ifelse(chow_test$p.value < 0.05, "Fixed Effect (FEM)", "Common Effect (CEM)"),
ifelse(hausman_test$p.value < 0.05, "Fixed Effect (FEM)", "Random Effect (REM)"),
ifelse(lm_test$p.value < 0.05, "Random Effect (REM)", "Common Effect (CEM)")
)
)
kable(tabel_pemilihan, row.names = FALSE, caption = "Tabel 6. Rangkuman Uji Pemilihan Model Terbaik")
Tabel 6. Rangkuman Uji Pemilihan Model Terbaik
| 1. Uji Chow (CEM vs FEM) |
60.8708 |
< 2e-16 |
Fixed Effect (FEM) |
| 2. Uji Hausman (REM vs FEM) |
6.6667 |
0.08332 |
Random Effect (REM) |
| 3. Uji Breusch-Pagan LM (CEM vs REM) |
1002.6720 |
< 2e-16 |
Random Effect (REM) |
cat("\n>>> KESIMPULAN MODEL TERPILIH OTOMATIS:", nama_terpilih, "<<<\n")
>>> KESIMPULAN MODEL TERPILIH OTOMATIS: Random Effect Model (REM) <<<
Uji Asumsi Klasik pada
Model Terpilih
# 1. Multikolinearitas (VIF)
ols_vif <- lm(model_formula, data = sampel_estimasi)
nilai_vif <- vif(ols_vif)
tabel_vif <- data.frame(
Variabel = names(nilai_vif),
VIF = round(nilai_vif, 4),
Keterangan = ifelse(nilai_vif < 10, "Bebas Multikolinearitas (VIF < 10)", "Terindikasi Multikolinearitas")
)
kable(tabel_vif, row.names = FALSE, caption = "Tabel 7. Hasil Uji Multikolinearitas (VIF)")
Tabel 7. Hasil Uji Multikolinearitas (VIF)
| X1 |
1.3207 |
Bebas Multikolinearitas (VIF < 10) |
| X2 |
1.1014 |
Bebas Multikolinearitas (VIF < 10) |
| X3 |
1.2215 |
Bebas Multikolinearitas (VIF < 10) |
# 2. Normalitas, Heteroskedastisitas, dan Autokorelasi
jb_test <- jarque.bera.test(residuals(model_terpilih))
bp_test <- bptest(model_formula, data = pdata, studentize = FALSE)
bg_test <- pbgtest(model_terpilih)
tabel_asumsi <- data.frame(
Jenis_Uji = c(
"Normalitas Residual (Jarque-Bera)",
"Heteroskedastisitas (Breusch-Pagan)",
"Autokorelasi (Breusch-Godfrey / Wooldridge)"
),
Statistik = round(c(jb_test$statistic, bp_test$statistic, bg_test$statistic), 4),
P_Value = format.pval(c(jb_test$p.value, bp_test$p.value, bg_test$p.value), digits = 4),
Status_Asumsi = ifelse(
c(jb_test$p.value, bp_test$p.value, bg_test$p.value) > 0.05,
"Terpenuhi (Lolos Uji)",
"Terlanggar (Perlu Estimasi Robust)"
)
)
kable(tabel_asumsi, row.names = FALSE, caption = paste("Tabel 8. Hasil Uji Asumsi Klasik pada", nama_terpilih))
Tabel 8. Hasil Uji Asumsi Klasik pada Random Effect Model
(REM)
| Normalitas Residual (Jarque-Bera) |
42.7918 |
5.104e-10 |
Terlanggar (Perlu Estimasi Robust) |
| Heteroskedastisitas (Breusch-Pagan) |
19.6975 |
0.0001961 |
Terlanggar (Perlu Estimasi Robust) |
| Autokorelasi (Breusch-Godfrey / Wooldridge) |
137.2654 |
< 2.2e-16 |
Terlanggar (Perlu Estimasi Robust) |
Estimasi Akhir dengan
Robust Standard Errors
# 1. Robust Arellano
rob_arellano <- coeftest(model_terpilih, vcov = vcovHC(model_terpilih, method = "arellano", type = "HC3"))
# 2. Robust Driscoll-Kraay
rob_dk <- coeftest(model_terpilih, vcov = vcovSCC(model_terpilih, type = "HC3"))
cat("=== ESTIMASI ROBUST ARELLANO (", nama_terpilih, ") ===\n", sep = "")
=== ESTIMASI ROBUST ARELLANO (Random Effect Model (REM)) ===
t test of coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 31.375516 3.386714 9.2643 < 2.2e-16 ***
X1 -0.167271 0.092563 -1.8071 0.0720537 .
X2 -0.148625 0.036652 -4.0551 6.863e-05 ***
X3 0.359311 0.098919 3.6324 0.0003463 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
cat("\n=== ESTIMASI ROBUST DRISCOLL-KRAAY (", nama_terpilih, ") ===\n", sep = "")
=== ESTIMASI ROBUST DRISCOLL-KRAAY (Random Effect Model (REM)) ===
t test of coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 31.375516 2.567349 12.2210 < 2.2e-16 ***
X1 -0.167271 0.048085 -3.4786 0.0006026 ***
X2 -0.148625 0.039344 -3.7776 0.0002017 ***
X3 0.359311 0.137795 2.6076 0.0097152 **
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
tabel_akhir <- data.frame(
Variabel = rownames(rob_dk),
Koefisien = round(rob_dk[, 1], 4),
Robust_Std_Error = round(rob_dk[, 2], 4),
T_Statistik = round(rob_dk[, 3], 4),
P_Value = round(rob_dk[, 4], 4),
Signifikansi = ifelse(rob_dk[, 4] < 0.01, "Signifikan (1%)",
ifelse(rob_dk[, 4] < 0.05, "Signifikan (5%)",
ifelse(rob_dk[, 4] < 0.10, "Signifikan (10%)", "Tidak Signifikan")))
)
kable(tabel_akhir, row.names = FALSE, caption = paste("Tabel 9. Hasil Akhir Regresi Robust Driscoll-Kraay -", nama_terpilih))
Tabel 9. Hasil Akhir Regresi Robust Driscoll-Kraay - Random
Effect Model (REM)
| (Intercept) |
31.3755 |
2.5673 |
12.2210 |
0.0000 |
Signifikan (1%) |
| X1 |
-0.1673 |
0.0481 |
-3.4786 |
0.0006 |
Signifikan (1%) |
| X2 |
-0.1486 |
0.0393 |
-3.7776 |
0.0002 |
Signifikan (1%) |
| X3 |
0.3593 |
0.1378 |
2.6076 |
0.0097 |
Signifikan (1%) |