r - Wu Hausman 检验来检验内生性

Question

我有一个模型，我怀疑内生性。我想通过 Wu hausman 测试来测试是否是这种情况，尽管我找不到任何方法来执行此操作。
已执行以下回归:

library(AER)
ivreg1 <- 
  ivreg(budget~ policyfactor+ control1 + control 2 | IV + control1 + control2, data=df)

df <- data.frame(
        budget = c(10, 8, -7, 8, 3, 2, 0.5, 1.5, -2, -15, 30, -0.5, 12, 18),
  policyfactor = c(1L, 1L, 0L, 0L, 0L, 1L, 1L, 0L, 0L, 0L, 1L, 0L, 1L, 1L),
            IV = c(17L,46L,18L,23L,35L,10L,0L,19L,
                   15L,12L,21L,6L,27L,10L),
      control1 = c(29.4,27.4,33.2,27.1,27.4,34.2,
                   26.8,32.9,26,26.3,27.3,33.4,23.5,31.3),
      control2 = c(1.3,20,-0.2,3,3.4,0.3,-1.1,1.9,
                   -2,-1.6,2.8,1.9,2,1.8)
)

创建于 2020-06-20 由 reprex package (v0.3.0)

Answer 1

Hausman 检验在 summary 中可用使用方法 diagnostics=TRUE选项。

summary(ivreg1, diagnostics=TRUE)
# Call:
#   AER::ivreg(formula = budget ~ policyfactor + control1 + control2 | 
#                IV + control1 + control2, data = df)
# 
# Residuals:
#   Min      1Q  Median      3Q     Max 
# -18.292  -4.920  -1.719   4.057  23.612 
# 
# Coefficients:
#              Estimate Std. Error t value Pr(>|t|)
# (Intercept)   14.4933    31.2550   0.464    0.653
# policyfactor   1.2168    15.1335   0.080    0.938
# control1      -0.3945     0.9835  -0.401    0.697
# control2       0.5168     0.7209   0.717    0.490
# 
# Diagnostic tests:
#                  df1 df2 statistic p-value
# Weak instruments   1  10     2.211   0.168
# Wu-Hausman         1   9     0.944   0.357            <--- THERE IT IS!
# Sargan             0  NA        NA      NA
# 
# Residual standard error: 11.46 on 10 degrees of freedom
# Multiple R-Squared: 0.1545,   Adjusted R-squared: -0.09916 
# Wald test: 0.381 on 3 and 10 DF,  p-value: 0.769