我在做一个时间序列分析,用Eviews估计出的方程中含有ar和ma项。所有的统计指标都很满意,但是ma的根有一个是-1.24,后面显示“Estimated MA process is noninvertible”。这是不是说明模型确认的有问题?对预测有什么影响?
Backcast: OFF (Roots of MA process too large)
Variable Coefficient Std. Error t-Statistic Prob.
AR(1) 0.101001 0.163438 0.617977 0.5396
AR(2) 0.314111 0.157705 1.991758 0.0522
AR(3) -0.575926 0.156795 -3.673127 0.0006
AR(4) 0.355170 0.144554 2.457000 0.0178
AR(5) 0.248594 0.129104 1.925531 0.0602
AR(6) -0.154357 0.118912 -1.298081 0.2006
MA(1) 0.217312 0.167282 1.299078 0.2003
MA(2) -0.338743 0.177864 -1.904504 0.0630
MA(3) 1.156392 0.165395 6.991690 0.0000
R-squared 0.521093 Mean dependent var 0.023683
Adjusted R-squared 0.439577 S.D. dependent var 0.118297
S.E. of regression 0.088559 Akaike info criterion -1.864073
Sum squared resid 0.368606 Schwarz criterion -1.538570
Log likelihood 61.19405 Durbin-Watson stat 1.843850
Inverted AR Roots .60 .55 .26-.83i .26+.83i
-.75 -.81
Inverted MA Roots .51+.82i .51-.82i -1.24
Estimated MA process is noninvertible


雷达卡



京公网安备 11010802022788号







