上图为S&P500从1990年到1995年的一周5日数据图,初看第一个图有趋势,所以ADF里加入trend, 看第三个图,first difference后好像是stationary了,检测数据如下,S&P500_1是Yt-1, DS&P500_1是∆Yt--1 ,从ADFADF-S&P500(ττ) = -1.545 大于1%的-3.969可以确证S&P500是有unit roots的non-stationary时间序列。从Akaike criterion (AIC)的值来看,lag(0)具有最小值,所以最终方程为∆Yt = α0+b0t+γYt-1+ εt 。
我现在有个疑问就是如何判断方程里要不要放trend,是通过t-value (2.0450)来看吗?这个值要和什么值比较?
Coefficient Std.Error t-value
S&P500_1 -0.0047733 0.0030901 -1.5447
Constant 1.4635 0.98821 1.4810
Trend 0.00096727 0.00047299 2.0450
DS&P500_1 0.029660 0.025455 1.1652
DS&P500_2 0.013615 0.025454 0.53489
DS&P500_3 -0.049539 0.025399 -1.9504
DS&P500_4 -0.0060462 0.025431 -0.23775
DS&P500_5 -0.0058873 0.025392 -0.23186
sigma = 2.83896 DW = 1.998 DW-S&P500 = 0.001677 ADF-S&P500(ττ) = -1.545
Critical values used in ADF test: 5%=-3.415, 1%=-3.969
RSS = 12500.62852 for 8 variables and 1559 observations
S&P500: ADF tests (T=1559, Constant+Trend; 5%=-3.42 1%=-3.97)
D-lag t-adf beta Y_1 sigma t-DY_lag t-prob AIC F-prob
5 -1.545 0.99523 2.839 -0.2319 0.8167 2.092
4 -1.568 0.99517 2.838 -0.2428 0.8082 2.091 0.8167
3 -1.591 0.99511 2.837 -1.958 0.0504 2.090 0.9452
2 -1.750 0.99464 2.840 0.5082 0.6114 2.091 0.2682
1 -1.715 0.99476 2.839 1.190 0.2343 2.090 0.3799
0 -1.627 0.99505 2.840 2.089 0.3458
Correlogram (ACF) and Portmanteau statistic (using TimeSeries90-95.xls)
The sample is: 1990-01-09 - 1995-12-29
S&P500 : Sample correlogram (ACF) from lag 1 to 36:
0.99648 0.99302 0.98952 0.98607 0.98265 0.97926
0.97595 0.97276 0.96948 0.96622 0.96267 0.95913
0.95533 0.95151 0.94763 0.94377 0.93997 0.93598
0.93215 0.92846 0.92489 0.92125 0.91759 0.91398
0.91054 0.90718 0.90394 0.90078 0.89757 0.89449
0.89126 0.88804 0.88484 0.88172 0.87841 0.87502
Partial autocorrelation function (PACF):
0.99648 0.0060142 -0.0077092 0.0060876 0.0030068 0.0017994
0.0094147 0.016523 -0.015665 0.0027536 -0.042864 -0.0019033
-0.037740 -0.0057484 -0.010164 -0.00029991 0.0054743 -0.031045
0.019615 0.017927 0.017073 -0.011747 -0.00066629 0.0049328
0.024072 0.012593 0.015562 0.012533 -0.011216 0.017231
-0.023968 6.3155e-005 0.00041851 0.0079040 -0.028972 -0.015562
Portmanteau statistic for 36 lags and 1565 observations: 49894.2
这个表中的假设H0:β= b0= α0=0,所用的Ø2统计怎么计算?我这个表中还有很多空白的部分,τατ 和τβτ我都找不到值,请高手帮忙。