用于面板 OLS 回归的 panelAR

Question

在使用 panelAR 函数执行面板数据回归时，设置 autoCorr = "none" 和 panelCorrMehotd = "none"，对标准误差执行了哪些具体校正？我想它会等于 OLS 回归，但它们略有不同。

require(panelAR)

data(Rehm)

tempFormula <- NURR ~ gini + selfemp + cum_right + tradeopen

outPanelAR <- panelAR(tempFormula, data=Rehm, panelVar = "ccode", timeVar = "year", 
               autoCorr="none",
               panelCorrMethod="none")

summary(outPanelAR)

Panel Regression with no autocorrelation and homoskedastic variance

Unbalanced Panel Design:                                             
 Total obs.:       75 Avg obs. per panel 3.75
 Number of panels: 20 Max obs. per panel 4   
 Number of times:  4  Min obs. per panel 1   

Coefficients:
              Estimate Std. Error t value Pr(>|t|)    
(Intercept) 109.498372   4.066262  26.929  < 2e-16 ***
gini         -1.987204   0.167623 -11.855  < 2e-16 ***
selfemp       0.288386   0.098038   2.942 0.004424 ** 
cum_right    -0.014059   0.002393  -5.875  1.3e-07 ***
tradeopen     0.080068   0.019576   4.090 0.000114 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

R-squared: 0.7123
Wald statistic: 185.7221, Pr(>Chisq(4)): 0

outLM <- lm(tempFormula, data=Rehm)

summary(outLM)

Call:
lm(formula = tempFormula, data = Rehm)

Residuals:
     Min       1Q   Median       3Q      Max 
-13.4097  -4.0381   0.3117   4.1815  13.8443 

Coefficients:
              Estimate Std. Error t value Pr(>|t|)    
(Intercept) 109.498372   4.208981  26.015  < 2e-16 ***
gini         -1.987204   0.173506 -11.453  < 2e-16 ***
selfemp       0.288386   0.101479   2.842 0.005872 ** 
cum_right    -0.014059   0.002477  -5.676 2.89e-07 ***
tradeopen     0.080068   0.020263   3.951 0.000183 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 6.054 on 70 degrees of freedom
Multiple R-squared:  0.7123,    Adjusted R-squared:  0.6959 
F-statistic: 43.34 on 4 and 70 DF,  p-value: < 2.2e-16

Answer 1

您需要将自由度校正设置为 TRUE 才能获得相同的结果：

outPanelAR2 <- panelAR(tempFormula, data=Rehm, panelVar = "ccode", timeVar = "year", 
                autoCorr="none",
                panelCorrMethod="none", 
                dof.correction = TRUE)
summary(outPanelAR2)

Panel Regression with no autocorrelation and homoskedastic variance

Unbalanced Panel Design:                                             
 Total obs.:       75 Avg obs. per panel 3.75
 Number of panels: 20 Max obs. per panel 4   
 Number of times:  4  Min obs. per panel 1   

Coefficients:
                           Estimate              Std. Error              t value   Pr(>|t|)    
(Intercept) 109.4983720122888115611   4.2089806065126253998  26.0154099999999993 < 2.22e-16 ***
gini         -1.9872037111259028830   0.1735058460730489749 -11.4532399999999992 < 2.22e-16 ***
selfemp       0.2883855296759073594   0.1014792385556597121   2.8418199999999998 0.00587203 ** 
cum_right    -0.0140594458309947108   0.0024770526774808253  -5.6758800000000003 2.8949e-07 ***
tradeopen     0.0800681086902131078   0.0202628791182291289   3.9514700000000000 0.00018318 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.050000000000000003 ‘.’ 0.10000000000000001 ‘ ’ 1

R-squared: 0.7123
Wald statistic: 173.3406, Pr(>Chisq(4)): 0

用于面板 OLS 回归的 panelAR

panelAR for panel OLS regression

r

panel-data

plm