With the knowledge of the sequential procedure for the unit root test, we would now test the data for R3M, R10Y, RGDP & M1/P with the Augmented Dicky-Fuller regression models of (model Ê with constant and trend), (model Ë with constant, no trend), and (model Ì with no constant and no trend).
From the SHAZAM output (see appendix 1), we can work through the following test with the respective data series:
R3M (raw data) Appropriate lag length suggested by SHAZAM is 2 for the analysis: |
|||
Test Model |
Test Type & Test Hypothesis |
Test Result |
Conclusion |
Ê |
F 3 test with H0: (a ,b ,r ) = (a ,0,0) against H1: (a ,b ,r ) ¹ (a ,0,0) | test statistic < critical value (Do not reject H0) |
Series is non-stationary, no trend with possible drift. |
Ê |
T-test H0: r =0 against H1:r ¹ 0 | t* < tcrit (Do not reject H0) |
Verified that series is non-stationary |
Ê |
F 2 test with H0: (a ,r ) = (0,0) against H1: (a ,r ) ¹ (0,0) | test statistic < critical value (Do not reject H0) |
Series is random walk (unit root, no trend) with no drift. |
Ë |
F 1 test with H0: (a ,b ,r ) = (0,0,0) against H1: (a ,b ,r ) ¹ (0,0,0) | test statistic < critical value (Do not reject H0) |
Series is non-stationary with no drift (no trend determined earlier) |
Ë |
T-test H0: r =0 against H1:r ¹ 0 | t* < tcrit (Do not reject H0) |
Verified that series is non-stationary (with restriction b =0 imposed) |
Ì |
T-test H0: r =0 against H1:r ¹ 0 | t* < tcrit (Do not reject H0) |
Verified that series is non-stationary (with restrictions a =b =0 imposed) |
Therefore, we can conclude that R3M is a random walk series (non-stationary, no trend) with zero drift. |
|||
R10Y (raw data) Appropriate lag length suggested by SHAZAM is 5 for the analysis: |
|||
Test Model |
Test Type & Test Hypothesis |
Test Result |
Conclusion |
Ê |
F 3 test with H0: (a ,b ,r ) = (a ,0,0) against H1: (a ,b ,r ) ¹ (a ,0,0) | test statistic < critical value (Do not reject H0) |
Series is non-stationary, no trend with possible drift. |
Ê |
T-test H0: r =0 against H1:r ¹ 0 | t* < tcrit (Do not reject H0) |
Verified that series is non-stationary |
Ê |
F 2 test with H0: (a ,r ) = (0,0) against H1: (a ,r ) ¹ (0,0) | test statistic < critical value (Do not reject H0) |
Series is random walk (unit root, no trend) with no drift. |
Ë |
F 1 test with H0: (a ,b ,r ) = (0,0,0) against H1: (a ,b ,r ) ¹ (0,0,0) | test statistic < critical value (Do not reject H0) |
Series is non-stationary with no drift (no trend determined earlier) |
Ë |
T-test H0: r =0 against H1:r ¹ 0 | t* < tcrit (Do not reject H0) |
Verified that series is non-stationary (with restriction b =0 imposed) |
Ì |
T-test H0: r =0 against H1:r ¹ 0 | t* < tcrit (Do not reject H0) |
Verified that series is non-stationary (with restrictions a =b =0 imposed) |
Therefore, we can conclude that R10Y is a random walk series (non-stationary, no trend) with zero drift. |
|||
RGDP (raw data) Appropriate lag length suggested by SHAZAM is 2 for the analysis: |
|||
Test Model |
Test Type & Test Hypothesis |
Test Result |
Conclusion |
Ê |
F 3 test with H0: (a ,b ,r ) = (a ,0,0) against H1: (a ,b ,r ) ¹ (a ,0,0) | test statistic < critical value (Do not reject H0) |
Series is non-stationary, no trend with possible drift. |
Ê |
T-test H0: r =0 against H1:r ¹ 0 | t* < tcrit (Do not reject H0) |
Verified that series is non-stationary |
Ê |
F 2 test with H0: (a ,r ) = (0,0) against H1: (a ,r ) ¹ (0,0) | test statistic > critical value (Reject H0) |
Series is random walk (unit root, no trend) with non-zero drift. |
Ë |
F 1 test with H0: (a ,b ,r ) = (0,0,0) against H1: (a ,b ,r ) ¹ (0,0,0) | test statistic > critical value (Reject H0) |
Series is non-stationary with non-zero drift (no trend determined earlier) |
Ë |
T-test H0: r =0 against H1:r ¹ 0 | t* < tcrit (Do not reject H0) |
Verified that series is non-stationary (with restriction b =0 imposed) |
Do not need to test model Ì as one of the restriction a =0 is not true and thus the result will be biased! |
|||
Therefore, we can conclude that RGDP is a random walk series (non-stationary, no trend) with a non-zero drift. |
|||
M1/P (raw data) Appropriate lag length suggested by SHAZAM is 5 for the analysis: |
|||
Test Model |
Test Type & Test Hypothesis |
Test Result |
Conclusion |
Ê |
F 3 test with H0: (a ,b ,r ) = (a ,0,0) against H1: (a ,b ,r ) ¹ (a ,0,0) | test statistic < critical value (Do not reject H0) |
Series is non-stationary, no trend with possible drift. |
Ê |
T-test H0: r =0 against H1:r ¹ 0 | t* < tcrit (Do not reject H0) |
Verified that series is non-stationary |
Ê |
F 2 test with H0: (a ,r ) = (0,0) against H1: (a ,r ) ¹ (0,0) | test statistic < critical value (Do not reject H0) |
Series is random walk (unit root, no trend) with no drift. |
Ë |
F 1 test with H0: (a ,b ,r ) = (0,0,0) against H1: (a ,b ,r ) ¹ (0,0,0) | test statistic < critical value (Do not reject H0) |
Series is non-stationary with no drift (no trend determined earlier) |
Ë |
T-test H0: r =0 against H1:r ¹ 0 | t* < tcrit (Do not reject H0) |
Verified that series is non-stationary (with restriction b =0 imposed) |
Ì |
T-test H0: r =0 against H1:r ¹ 0 | t* < tcrit (Do not reject H0) |
Verified that series is non-stationary (with restrictions a =b =0 imposed) |
Therefore, we can conclude that M1/P is a random walk series (non-stationary, no trend) with zero drift. |
|||
R3M (First-order differenced data) Appropriate lag length suggested by SHAZAM is 8 for the analysis: |
|||
Test Model |
Test Type & Test Hypothesis |
Test Result |
Conclusion |
Ê |
F 3 test with H0: (a ,b ,r ) = (a ,0,0) against H1: (a ,b ,r ) ¹ (a ,0,0) | test statistic > critical value (Reject H0) |
Could be either or or |
Ê |
T-test H0: r =0 against H1:r ¹ 0 | t* > tcrit (Reject H0) |
Could be either or , ie. no unit root. |
Í * |
T-test H0: b =0 against H1:b ¹ 0 | t* < tcrit (Do not reject H0) |
Series is stationary (no unit root) with no trend. |
Í * |
T-test H0: a =0 against H1:a ¹ 0 | t* < tcrit (Do not reject H0) |
Series is stationary with no trend and zero (no) drift. |
Í * |
T-test H0: r =0 against H1:r ¹ 0 | t* > tcrit (Reject H0) |
Verified that series is stationary (no unit root). |
Ï * |
T-test H0: r =0 against H1:r ¹ 0 | t* > tcrit (Reject H0) |
Verified that series is stationary (with restrictions a =b =0 imposed). |
Therefore, we can conclude that the first-order differenced R3M is a stationary series (no unit root) with no trend and zero drift. | |||
R10Y (First-order differenced data) Appropriate lag length suggested by SHAZAM is 8 for the analysis: |
|||
Test Model |
Test Type & Test Hypothesis |
Test Result |
Conclusion |
Ê |
F 3 test with H0: (a ,b ,r ) = (a ,0,0) against H1: (a ,b ,r ) ¹ (a ,0,0) | test statistic > critical value (Reject H0) |
Could be either or or |
Ê |
T-test H0: r =0 against H1:r ¹ 0 | t* > tcrit (Reject H0) |
Could be either or , ie. no unit root. |
Í * |
T-test H0: b =0 against H1:b ¹ 0 | t* < tcrit (Do not reject H0) |
Series is stationary (no unit root) with no trend. |
Í * |
T-test H0: a =0 against H1:a ¹ 0 | t* < tcrit (Do not reject H0) |
Series is stationary with no trend and zero (no) drift. |
Í * |
T-test H0: r =0 against H1:r ¹ 0 | t* > tcrit (Reject H0) |
Verified that series is stationary (no unit root). |
Ï * |
T-test H0: r =0 against H1:r ¹ 0 | t* > tcrit (Reject H0) |
Verified that series is stationary (with restrictions a =b =0 imposed). |
Therefore, we can conclude that the first-order differenced R10Y is a stationary series (no unit root) with no trend and zero drift. | |||
RGDP (First-order differenced data) Appropriate lag length suggested by SHAZAM is 9 for the analysis: |
|||
Test Model |
Test Type & Test Hypothesis |
Test Result |
Conclusion |
Ê |
F 3 test with H0: (a ,b ,r ) = (a ,0,0) against H1: (a ,b ,r ) ¹ (a ,0,0) | test statistic > critical value (Reject H0) |
Could be either or or |
Ê |
T-test H0: r =0 against H1:r ¹ 0 | t* > tcrit (Reject H0) |
Could be either or , ie. no unit root. |
Í * |
T-test H0: b =0 against H1:b ¹ 0 | t* < tcrit (Do not reject H0) |
Series is stationary (no unit root) with no trend. |
Í * |
T-test H0: a =0 against H1:a ¹ 0 | t* > tcrit (Reject H0) |
Series is stationary with no trend and a non-zero drift. |
Í * |
T-test H0: r =0 against H1:r ¹ 0 | t* > tcrit (Reject H0) |
Verified that series is stationary (no unit root). |
Î * |
T-test H0: a =0 against H1:a ¹ 0 | t* > tcrit (Reject H0) |
Verified that series has a drift term (with restriction b =0 imposed). |
Î * |
T-test H0: r =0 against H1:r ¹ 0 | t* > tcrit (Reject H0) |
Verified that series is stationary (with restriction b =0 imposed). |
Therefore, we can conclude that the first-order differenced RGDP is a stationary series (no unit root) with no trend and a non-zero drift. | |||
M1/P (First-order differenced data) Appropriate lag length suggested by SHAZAM is 4 for the analysis: |
|||
Test Model |
Test Type & Test Hypothesis |
Test Result |
Conclusion |
Ê |
F 3 test with H0: (a ,b ,r ) = (a ,0,0) against H1: (a ,b ,r ) ¹ (a ,0,0) | test statistic > critical value (Reject H0) |
Could be either or or |
Ê |
T-test H0: r =0 against H1:r ¹ 0 | t* > tcrit (Reject H0) |
Could be either or , ie. no unit root. |
Í * |
T-test H0: b =0 against H1:b ¹ 0 | t* < tcrit (Do not reject H0) |
Series is stationary (no unit root) with no trend. |
Í * |
T-test H0: a =0 against H1:a ¹ 0 | t* < tcrit (Do not reject H0) |
Series is stationary with no trend and zero (no) drift. |
Í * |
T-test H0: r =0 against H1:r ¹ 0 | t* > tcrit (Reject H0) |
Verified that series is stationary (no unit root). |
Ï * |
T-test H0: r =0 against H1:r ¹ 0 | t* > tcrit (Reject H0) |
Verified that series is stationary (with restrictions a =b =0 imposed). |
Therefore, we can conclude that the first-order differenced M1/P is a stationary series (no unit root) with no trend and zero drift. |
*Note: Model Í is from Appendix 2, or from Appendix 3.
Model Î is from Appendix 2, or from Appendix 3.
Model Ï is from Appendix 2, or from Appendix 3.
Thus, it was illustrated that the problem of non-stationarity of the raw data set of R3M, R10Y, RGDP, M1/P can be removed by taking first-order difference of the respective data. The unit root test has been used to verify the transition of non-stationary series of the raw data to a stationary one following first-order differencing at I(1).
Return to Assignments for Semester 1, 1998