Deal allstaters,
I am examining a sequence which has passed tests for normality (Jarque-Bera,
Kolmogorov-Smirnov) and RUNS test for independence. When I try an identification
approach, I find that there are significant autocorrelation and partial autocorrelation
coefficients for the AC an PAC functions as seen below from the Q-statistic. Is that
possible and under which circumstances ? Any references extremely welcome. This
sequence were residuals obtained by wavelet denoising of the FTSE ALL SHARE
index returns via the "waveshrink" algorithm.
Lag AC PAC Q-Stat Prob
1 0.023 0.023 4.3134 0.038
2 -0.096 -0.097 79.909 0.000
3 -0.104 -0.100 168.70 0.000
4 -0.084 -0.091 226.15 0.000
5 -0.066 -0.086 262.19 0.000
6 -0.071 -0.102 303.63 0.000
7 -0.099 -0.141 384.48 0.000
8 -0.031 -0.084 392.25 0.000
9 -0.011 -0.084 393.30 0.000
10 0.015 -0.061 395.13 0.000
Thanks in advance for your time.
Best regards,
Costas
Costas Vorlow, Economics & Finance
University of Durham,
23/26 Old Elvet,
Durham, DH13HY,
UK
http://www.dur.ac.uk/k.e.vorloou
http://www.dur.ac.uk/Economics
|