EQC7006 : Time Series Analysis 2019
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
ACADEMIC SESSION 2018/2019: SEMESTER II
EQC7006: Time Series Analysis
1. Answer the following questions.
a) Given below is an annual data for 12 years. Estimate trend-cycle component of the series using a single 3-period moving average (3 MA) and a 3-term weighted moving average (3 WMA) with weights of 0.25, 0.5, and 0.25.
(4 marks)
Write formulae used to estimate irregular component ( ) for an annual time series data (additive and multiplicative trend).
(1 mark) (Total: 5 marks)
2. The following tables give estimated coefficients for three regression models. Y denotes the monthly temperature (in Celsius) series. T represents time trend (T = 1, 2, 3, …, 192). Di’s are seasonal dummies where D1=1 for January and 0 otherwise, and D2, D3, D4, D5, D6, D7, D8, D9, D10, D11 and D12 are similarly defined for February, March until December, respectively.
Model 1
Dependent Variable: Y
Method: Least Squares
Sample: 2000M01 2015M12
Included observations: 192
Model 2
Dependent Variable: Y
Method: Least Squares
Sample: 2000M01 2015M12
Included observations: 192
a) What estimates are given by the twelve coefficients of Model 1? (1 mark)
b) In between Model 1 and Model 2, which one should be used? Why? (2 marks)
c) Based on Model 1, perform a seasonal variation test at 5% significance level. State the null and alternative hypotheses of the test. (4 marks)
d) Given below are out-sample (January 2016 – December 2016) forecast accuracy measures produced by Model 1. Interpret test set MAPE, ACF1 and Theil’s U values.
3. The following time series plot represents monthly data of rainfall amount (in millimeters) in Malaysia for the period of 16 years (January 2000 – December 2015).
a) Evaluate and discuss the results of ADF and KPSS tests given below. Given that seasonal variation is present, decide whether seasonal differencing is needed before any ARMA or ARIMA model can be estimated for the above series.
Null Hypothesis: RAINFALL has a unit root
Exogenous: Constant
Lag Length: 0 (Automatic - based on SIC, maxlag=14)
|
t-Statistic |
Prob.* |
|
Augmented Dickey-Fuller test statistic |
-8.939625 |
0.0000 |
|
Test critical values: |
1% level |
-3.464460 |
|
|
5% level |
-2.876435 |
|
|
10% level |
-2.574788 |
|
*MacKinnon (1996) one-sided p-values.
Null Hypothesis: RAINFALL has a unit root
Exogenous: Constant, Linear Trend
Lag Length: 0 (Automatic - based on SIC, maxlag=14)
t-Statistic Prob.*
|
||
Augmented Dickey-Fuller test statistic -8.919751 0.0000 |
||
Test critical values: |
1% level 5% level 10% level |
-4.006566 -3.433401 -3.140550 |
|
*MacKinnon (1996) one-sided p-values.
Null Hypothesis: RAINFALL is stationary
Exogenous: Constant
Bandwidth: 1 (Newey-West automatic) using Bartlett kernel
LM-Stat. |
|
Kwiatkowski-Phillips-Schmidt-Shin test statistic |
0.155087 |
Asymptotic critical values*: 1% level |
0.739000 |
5% level |
0.463000 |
10% level |
0.347000 |
*Kwiatkowski-Phillips-Schmidt-Shin (1992, Table 1)
Null Hypothesis: RAINFALL is stationary
Exogenous: Constant, Linear Trend
Bandwidth: 1 (Newey-West automatic) using Bartlett kernel
LM-Stat. |
|
Kwiatkowski-Phillips-Schmidt-Shin test statistic |
0.135448 |
Asymptotic critical values*: 1% level |
0.216000 |
5% level |
0.146000 |
10% level |
0.119000 |
*Kwiatkowski-Phillips-Schmidt-Shin (1992, Table 1)
(2 marks)
b) Inspect correlograms of the series given below. Based on your finding in (a), propose two ARMA or ARIMA models for the given series and give justifications.
* Partial autocorrelations beyond lag 36 are all insignificant.
* Partial autocorrelations beyond lag 36 are all insignificant.
(5 marks)
c) By looking at the time plot of the series, can you decide whether the proposed models should include a constant term or not?
(0.5 mark) (Total: 7.5 marks)
4. The following time series plot represents monthly data of temperature (in Celcius) in Malaysia for the period of 16 years (January 2000 – December 2015).
a) Evaluate and discuss the results of ADF and KPSS tests given below. Assuming that seasonal variation is present, decide whether seasonal differencing is needed before any ARMA or ARIMA model can be estimated for the above series.
Null Hypothesis: TEMPERATURE has a unit root
Exogenous: Constant
Lag Length: 13 (Automatic - based on SIC, maxlag=14)
|
t-Statistic |
Prob.* |
|
Augmented Dickey-Fuller test statistic |
-2.683750 |
0.0787 |
|
Test critical values: |
1% level |
-3.464460 |
|
|
5% level |
-2.876435 |
|
|
10% level |
-2.574788 |
|
*MacKinnon (1996) one-sided p-values.
Null Hypothesis: TEMPERATURE has a unit root
Exogenous: Constant, Linear Trend
Lag Length: 13 (Automatic - based on SIC, maxlag=14)
t-Statistic Prob.*
|
||
Augmented Dickey-Fuller test statistic -2.714042 0.2322 |
||
Test critical values: |
1% level 5% level 10% level |
-4.006566 -3.433401 -3.140550 |
|
*MacKinnon (1996) one-sided p-values.
Null Hypothesis: TEMPERATURE is stationary
Exogenous: Constant
Bandwidth: 4 (Newey-West automatic) using Bartlett kernel
LM-Stat. |
|
Kwiatkowski-Phillips-Schmidt-Shin test statistic |
0.141894 |
Asymptotic critical values*: 1% level |
0.739000 |
5% level |
0.463000 |
10% level |
0.347000 |
*Kwiatkowski-Phillips-Schmidt-Shin (1992, Table 1)
Null Hypothesis: TEMPERATURE is stationary
Exogenous: Constant, Linear Trend
Bandwidth: 4 (Newey-West automatic) using Bartlett kernel
LM-Stat. |
|
Kwiatkowski-Phillips-Schmidt-Shin test statistic |
0.130433 |
Asymptotic critical values*: 1% level |
0.216000 |
5% level |
0.146000 |
10% level |
0.119000 |
*Kwiatkowski-Phillips-Schmidt-Shin (1992, Table 1) (2 marks)
b) Inspect correlograms of the series given below. Based on your finding in (a), propose two ARMA or ARIMA models for the given series and give justifications.
* Partial autocorrelations beyond lag 36 are all insignificant.
* Partial autocorrelations beyond lag 36 are all insignificant. (5 marks)
c) By looking at the time plot of the series, can you decide whether the proposed models should include a constant term or not?
(0.5 mark) (Total: 7.5 marks)
5. Given below is estimation results of an ARIMA model for a monthly time series data, Y.
Dependent Variable: D(Y,0,12)
Method: ARMA Conditional Least Squares (Marquardt - EViews legacy)
Sample: 2000M01 2015M12
Included observations: 192
a) Does the estimated process is covariance stationary? Explain. (1 mark)
b) Does the estimated process is invertible? Explain. (1 mark)
c) Write the model in terms of the backshift operator and without backshift operator. (3 marks)
d) Table A below gives the last 24 observations of the series (January 2014 to December 2015). Table B on the other hand gives the last 24 residuals generated by the model. Compute forecasts for January and February of 2016 and 95% forecast intervals (given that forecast error variance: January 2016=0.0663, February 2016=0.0754).
2022-06-15