本文介绍了如何设定一个涉及时间序列数据的回归模型，包括处理随机误差项的异方差性和自相关性，以及在R语言中通过`reghdf`函数进行估计的实现步骤，如生成虚拟变量、设置面板数据集和进行空间与时间自相关性检验。
1.模型设定
Y i , t = β 0 + β 1 ⋅ X i , t + γ i + γ t + ϵ i , t a ( 1 ) Y_{i,t}=\beta_{0} + \beta_{1} \cdot X_{i,t}+\gamma_{i} + \gamma_{t} + \epsilon_{i,t}^{a} \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ (1) Yi,t=β0+β1⋅Xi,t+γi+γt+ϵi,ta (1)
Y i , t − 1 = β 0 + β 1 ⋅ X i , t − 1 + γ i + γ t − 1 + ϵ i , t − 1 b ( 2 ) Y_{i,t-1}=\beta_{0} + \beta_{1} \cdot X_{i,t-1}+\gamma_{i} + \gamma_{t-1} + \epsilon_{i,t-1}^{b} \ \ \ \ \ \ \ \ \ \ \ (2) Yi,t−1=β0+β1⋅Xi,t−1+γi+γt−1+ϵi,t−1b (2)
(1)式 - (2)式：
Y
i
,
t
−
Y
i
,
t
−
1
=
(
β
0
−
β
0
)
+
β
1
(
X
i
,
t
−
X
i
,
t
−
1
)
+
(
γ
t
−
γ
t
−
1
)
+
(
ϵ
i
,
t
a
−
ϵ
i
,
t
b
)
Y_{i,t} - Y_{i,t-1} =(\beta_{0} -\beta_{0})+ \beta_{1}(X_{i,t}-X_{i,t-1})+(\gamma_{t}-\gamma_{t-1}) +(\epsilon_{i,t}^{a}-\epsilon_{i,t}^{b})
Yi,t−Yi,t−1=(β0−β0)+β1(Xi,t−Xi,t−1)+(γt−γt−1)+(ϵi,ta−ϵi,tb)
进而得到：
Δ
Y
i
,
t
=
β
1
⋅
Δ
X
i
,
t
+
σ
t
+
ϵ
i
,
t
\Delta Y_{i,t}=\beta1 \cdot \Delta X_{i,t} + \sigma_{t} + \epsilon_{i,t}
ΔYi,t=β1⋅ΔXi,t+σt+ϵi,t
2.实现过程
clear
set obs 1000
egen citycode = repeat() ,v(1000(10)2000)
gen provcode = real(substr(string(citycode),1,2))
sort citycode
egen year = repeat(), v(2005/2015)
gen e = rnormal()
gen cov1 = 2*e + 6
gen cov2 = -e*cov1 + 10
gen x = uniform()* 3
gen iv =
gen y = 2*x + cov1 + cov2 + citycode / 1000 + (year-2004)*5*uniform() // 系数应当为2
reghdfe y x cov1 cov2 ,a(citycode year)
reghdfe y x ,a(provcode year)
xtset citycode year
gen dx = d.x
gen dy = d.y
reghdfe dy dx ,a(citycode)
扫一扫
9387
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
