面板单位根检验(Stata)
对于包含时间序列较长的面板数据,在回归前需要进行单位根检验,否则可能存在虚假回归问题。检验方法在Stata使用手册中可以查阅得到,包括LLC检验、HT检验、Breitung检验、IPS检验、fisher 检验以及Hardri LM检验。下面具体介绍相关命令及使用方法。
**# versions 17.0
**# 1 数据准备
webuse pennxrate,clear // 调用网页数据
describe //描述统计
*=====================================
**# 2 LLC检验
/*
note:
仅适用于长面板数据,即时间长度远大于截面个数
假定不存在截面相关,否则存在显著性扭曲,需要先去截面均值
允许不同截面内自回归系数相同
必须是平衡面板
*/
xtunitroot llc lnrxrate , demean lags(aic 10) kernel(bartlett nwest)
/*
option:
demean表示去截面均值
lags(#) 表示序列变量差分的滞后项数#,其中截面滞后阶数相同
lags(aic #) lags(bic #) lags(hqic #) 表示分别以aic bic hqic准则判定最大滞后阶数#
trend 表示加入趋势项并默认加入个体固定项
noconstant 表示趋势项与个体项都不加入
trend和noconstant都不加默认个体固定项
kernel(kernel_spec) 为核函数,估计渐进方差,具体设定包括 ba pa qu等)
结果输出:
Levin–Lin–Chu unit-root test for lnrxrate
-----------------------------------------
H0: Panels contain unit roots Number of panels = 151
Ha: Panels are stationary Number of periods = 34
AR parameter: Common Asymptotics: N/T -> 0
Panel means: Included
Time trend: Not included Cross-sectional means removed
ADF regressions: 2.18 lags average (chosen by AIC)
LR variance: Bartlett kernel, 3.50 lags average
(chosen by Newey–West)
------------------------------------------------------------------------------
Statistic p-value
------------------------------------------------------------------------------
Unadjusted t -22.1334
Adjusted t* -1.8763 0.0303
------------------------------------------------------------------------------
调整t统计量为 -1.8763 ,伴随概率 0.0303<0.05,拒绝H0: Panels contain unit roots
*/
*=====================================
**# 3 HT 检验
/*
note:
适合于时间维度较短的微观面板数据
允许不同截面内自回归系数相同
必须是平衡面板
*/
xtunitroot ht lnrxrate , demean
/*
option:
trend 表示加入趋势项并默认加入个体固定项
noconstant 表示趋势项与个体项都不加入
trend和noconstant都不加默认个体固定项
demean表示去截面均值
robust:允许截面相关
lags(#) 预白噪声化 #,其中截面滞后阶数相同
结果输出:Harris-Tzavalis unit-root test for lnrxrate
-------------------------------------------
H0: Panels contain unit roots Number of panels = 151
Ha: Panels are stationary Number of periods = 34
AR parameter: Common Asymptotics: N -> Infinity
Panel means: Included T Fixed
Time trend: Not included Cross-sectional means removed
------------------------------------------------------------------------------
Statistic z p-value
------------------------------------------------------------------------------
rho 0.8184 -13.1239 0.0000
------------------------------------------------------------------------------
z = -13.1239 ,伴随概率 0.000<0.01,拒绝H0: Panels contain unit roots
*/
*=====================================
**# 4 Breitung 检验
/*
option:
1)允许不同截面内自回归系数相同
2)必须是平衡面板
*/
xtunitroot breitung lnrxrate if g7, lags(3) robust
/*
option:
trend:加入时间趋势
noconstant : 趋势项与个体项都不加入
demean :去截面均值
robust:允许截面相关
lags(#) :预白噪声数
结果:
Breitung unit-root test for lnrxrate
--------------------------------------
H0: Panels contain unit roots Number of panels = 6
Ha: Panels are stationary Number of periods = 34
AR parameter: Common Asymptotics: T,N -> Infinity
Panel means: Included sequentially
Time trend: Not included Prewhitening: 3 lags
------------------------------------------------------------------------------
Statistic p-value
------------------------------------------------------------------------------
lambda* -1.2258 0.1101
------------------------------------------------------------------------------
* Lambda robust to cross-sectional correlation
统计量 lambda* = -1.2258 伴随概率 0.1101>0.1,不拒绝H0: Panels contain unit roots
*/
*=====================================
**# 5 IPS检验
/*
不同截面内自回归系数不同
允许非平衡
*/
xtunitroot ips lnrxrate, lags(aic 5)
/*
option:
trend :包含时间趋势
demean :取截面均值
lags(#) :为ADF回归设定滞后结构
结果输出:
Im–Pesaran–Shin unit-root test for lnrxrate
-------------------------------------------
H0: All panels contain unit roots Number of panels = 151
Ha: Some panels are stationary Number of periods = 34
AR parameter: Panel-specific Asymptotics: T,N -> Infinity
Panel means: Included sequentially
Time trend: Not included
ADF regressions: 1.11 lags average (chosen by AIC)
------------------------------------------------------------------------------
Statistic p-value
------------------------------------------------------------------------------
W-t-bar -15.2812 0.0000
------------------------------------------------------------------------------
Wt统计量 = -15.2812 伴随概率0.000<0.01,拒绝H0: All panels contain unit roots
*/
*=====================================
**# 6 fisher 检验
/*
不同截面内自回归系数不同
允许非平衡
*/
xtunitroot fisher lnrxrate, dfuller lags(3) drift
/*
option:
* dfuller :使用 ADF 单位根检验
* pperron :使用pp单位根检验
* lags(#) :为ADF回归设定滞后结构
demean 同上文
Fisher-type unit-root test for lnrxrate
Based on augmented Dickey–Fuller tests
---------------------------------------
H0: All panels contain unit roots Number of panels = 151
Ha: At least one panel is stationary Number of periods = 34
AR parameter: Panel-specific Asymptotics: T -> Infinity
Panel means: Included
Time trend: Not included
Drift term: Included ADF regressions: 3 lags
------------------------------------------------------------------------------
Statistic p-value
------------------------------------------------------------------------------
Inverse chi-squared(302) P 916.1451 0.0000
Inverse normal Z -18.8512 0.0000
Inverse logit t(759) L* -19.5571 0.0000
Modified inv. chi-squared Pm 24.9892 0.0000
------------------------------------------------------------------------------
P statistic requires number of panels to be finite.
Other statistics are suitable for finite or infinite number of panels.
------------------------------------------------------------------------------
上述逆卡方、逆正态、逆逻辑以及修正逆卡方统计量伴随概率均为0.000<0.01,故
拒绝H0: All panels contain unit roots
*/
*=====================================
**# 7 Hardri LM检验
xtunitroot hadri lnrxrate, kernel(parzen 5)
/*
option作用同上文相关检验
trend include a time trend
demean subtract cross-sectional means
robust allow for cross-sectional dependence
kernel(kernel_spec) specify method to estimate long-run varianc
*/
更多内容help xtunitroot