在面板数据中,我用dlog()和d()进行OLS回归,因为在panel data 的单位根检验中有好几组变量不平稳,想采用一阶差分去分析,如下所示:
Dependent Variable: DLOG(AGRIAREA?)
Method: Pooled EGLS (Cross-section weights)
Linear estimation after one-step weighting matrix
White cross-section standard errors & covariance (d.f. corrected)
Variable Coefficient Std. Error t-Statistic Prob.
C -0.016555 0.001016 -16.30031 0.0000
DLOG(HHAREA?) -0.036204 0.006837 -5.295425 0.0000
D(AGSHR?) -0.735842 0.100036 -7.355782 0.0000
DLOG(EMPSHR?) 0.121324 0.016958 7.154149 0.0000
D(RI?) -0.005669 0.001113 -5.094309 0.0000
DLOG(PS11?(-1)) -0.007268 0.000360 -20.17145 0.0000
D(RI?*LOG(PS11?(-1))) 0.000689 9.21E-05 7.486651 0.0000
Weighted Statistics
R-squared 0.921061 Mean dependent var -0.035767
Adjusted R-squared 0.870315 S.D. dependent var 0.091151
S.E. of regression 0.020465 Sum squared resid 0.023453
F-statistic 18.15033 Durbin-Watson stat 2.729444
Prob(F-statistic) 0.000000
不知道这样的结果该怎么样去解释经济现象,两期的差值,想不通该怎么去解释……
还请大家多多关照,多多出主意阿!
(单位根、自相关与协整,是不是要在回归之前都要做呢?在eviews5.1中,面板数据只有unit root test可以直接做,correlogram和cointegration只能对残差去做,不是太懂原理,请大家指点指点。)
3q