ISYE 6402 Time Series Analysis Final Exam Prep Document |
2026/2027 Edition | 250 Verified Questions
ISYE 6402 Final Exam 2026-2027 QUESTIONS AND ANSWERS ALREADY GRADED A+. 100% Verified
Solutions | Updated Per Latest Guidelines | Graded A+
This comprehensive exam preparation resource contains 250 verified questions and detailed rationales
for the ISYE 6402 Time Series Analysis final exam. Each question is aligned with the latest 2026/2027
curriculum, covering all major topics from stationarity and ARIMA models to spectral analysis and
forecasting. The document is designed to ensure a thorough understanding of time series concepts and
to guarantee a passing grade with A+ performance. Every answer includes step-by-step explanations
and distractor analyses to reinforce learning.
Abstract:
This exam preparation document for ISYE 6402 Time Series Analysis provides a rigorous and comprehensive
review of all core concepts required for the final exam. The material is organized into content areas that mirror the
course syllabus, ensuring systematic coverage. Each of the 250 verified questions is accompanied by a detailed
rationale that explains the correct answer and analyzes common distractors, thereby deepening conceptual
understanding. The document emphasizes practical application through real-world data examples and forecasting
exercises. Special attention is given to model identification, estimation, and diagnostic checking, as well as
advanced topics such as spectral analysis and cointegration. By working through these questions, students will
develop the analytical skills necessary to excel in the exam and apply time series methods in research or industry.
The content has been updated to reflect the latest academic year and incorporates feedback from previous exam
cycles. This resource is designed to be the sole study aid needed to achieve a top grade.
Content Area Overview:
Content Area Questions Key Topics Weight
Stationarity and Time Series 1-40 Stationarity, autocorrelation, partial 16%
Basics autocorrelation, white noise, transformations
ARMA and ARIMA Models 41-90 AR, MA, ARMA, ARIMA, identification, 20%
estimation, diagnostic checking
Seasonal Models and SARIMA 91-130 Seasonal differencing, SARIMA, seasonal 16%
decomposition, model selection
Forecasting 131-170 Point forecasts, prediction intervals, forecast 16%
accuracy measures, rolling forecasts
Spectral Analysis 171-210 Periodogram, spectral density, filtering, 16%
frequency domain analysis
Advanced Topics 211-250 GARCH, state-space models, cointegration, 16%
multivariate time series
Page 1
,Q1. Given a time series with a unit root, which of the following transformations is most appropriate
to achieve stationarity?
A. First differencing
B. Log transformation
C. Seasonal differencing
D. Box-Cox transformation
Correct Answer: A. First differencing
Rationale: A unit root implies non-stationarity that can be removed by first differencing, which
transforms the series to its changes. Log transformation stabilizes variance but does not remove unit
roots. Seasonal differencing addresses seasonal unit roots. Box-Cox transformation is for variance
stabilization.
Why Wrong:
B - Log transformation stabilizes variance but does not remove unit roots.
C - Seasonal differencing is for seasonal unit roots, not a simple unit root.
D - Box-Cox transformation is for variance stabilization, not for removing unit roots.
Reference: Brockwell & Davis (2016). Introduction to Time Series and Forecasting, 3rd Ed., Ch. 5
Q2. In spectral analysis, the periodogram is defined as the squared magnitude of the Fourier
transform of the data. Which of the following is a key limitation of the periodogram as an estimator
of the spectral density?
A. It is biased but consistent.
B. It is unbiased but inconsistent.
C. It is both biased and inconsistent.
D. It is asymptotically unbiased and consistent.
Correct Answer: B. It is unbiased but inconsistent.
Rationale: The periodogram is an unbiased estimator of the spectral density for a given frequency, but its
variance does not decrease with sample size, making it inconsistent. Smoothing methods are needed to
achieve consistency.
Why Wrong:
A - The periodogram is unbiased, not biased.
C - The periodogram is unbiased, but it is inconsistent.
D - The periodogram is unbiased but inconsistent, not consistent.
Reference: Shumway & Stoffer (2017). Time Series Analysis and Its Applications, 4th Ed., Ch. 4
Page 2
,Q3. Consider a state space model with observation equation y_t = _t + _t and state equation _t =
_{t-1} + _t, where _t and _t are independent white noise processes with variances _^2 and _^2
respectively. This model is equivalent to which of the following?
A. ARIMA(0,1,1) model
B. ARIMA(1,0,0) model
C. ARIMA(0,0,1) model
D. ARIMA(1,1,0) model
Correct Answer: A. ARIMA(0,1,1) model
Rationale: This is the local level model, which yields an ARIMA(0,1,1) representation for y_t. The first
difference of y_t is an MA(1) process with parameter depending on the signal-to-noise ratio.
Why Wrong:
B - ARIMA(1,0,0) is an AR(1) model, which does not have a unit root.
C - ARIMA(0,0,1) is an MA(1) model for the level, not differenced.
D - ARIMA(1,1,0) is an ARI(1,1) model, not equivalent to the local level.
Reference: Durbin & Koopman (2012). Time Series Analysis by State Space Methods, 2nd Ed., Ch. 2
Q4. Two nonstationary time series, X_t and Y_t, are cointegrated if there exists a linear combination
Z_t = Y_t - X_t that is stationary. Which of the following statements about cointegration is correct?
A. Both series must be integrated of order 1, and the cointegrating vector is unique.
B. Both series must be integrated of the same order, and the cointegrating vector is unique up to scalar
multiplication.
C. Both series must be integrated of order 1, and there can be multiple cointegrating vectors if there
are more than two series.
D. Both series must be stationary, and cointegration is equivalent to correlation.
Correct Answer: B. Both series must be integrated of the same order, and the cointegrating vector is
unique up to scalar multiplication.
Rationale: For cointegration, each series must be integrated of the same order d (typically d=1). The
cointegrating vector is unique up to a scalar multiple when there are two series. With more than two
series, multiple cointegrating vectors may exist.
Why Wrong:
A - The order of integration need not be exactly 1; but it must be the same for both.
C - For two series, the cointegrating vector is unique up to scaling, not multiple.
D - Cointegration requires nonstationary series, not stationary.
Reference: Hamilton (1994). Time Series Analysis, Ch. 19
Page 3
, Q5. In a GARCH(1,1) model, the unconditional variance of the error term exists if and only if:
A. 1 + 1 < 1
B. 1 + 1 1
C. 1 + 1 > 1
D. 1 + 1 = 1
Correct Answer: A. 1 + 1 < 1
Rationale: For a GARCH(1,1) model, the unconditional variance is finite and constant if ±1 + ²1 < 1. If
1 + 1 = 1, the process is IGARCH and the unconditional variance is infinite.
Why Wrong:
B - When 1+1=1, unconditional variance is infinite (IGARCH).
C - If 1+1>1, the variance explodes over time.
D - Equality to 1 corresponds to IGARCH, not finite variance.
Reference: Tsay (2010). Analysis of Financial Time Series, 3rd Ed., Ch. 3
Q6. A researcher fits an ARMA(2,2) model to a time series and obtains residuals. The Ljung-Box
test on the residuals yields a p-value of 0.03 at lag 10. Which of the following is the most appropriate
conclusion at the 5% significance level?
A. The residuals are white noise; the model is adequate.
B. There is evidence of remaining autocorrelation in the residuals; the model may be inadequate.
C. The test is inconclusive because the p-value is close to 0.05.
D. The residuals are normally distributed.
Correct Answer: B. There is evidence of remaining autocorrelation in the residuals; the model may
be inadequate.
Rationale: A p-value of 0.03 < 0.05 indicates significant autocorrelation in the residuals at lag 10,
suggesting the model does not capture all temporal dependencies. The Ljung-Box test does not test
normality.
Why Wrong:
A - A p-value less than 0.05 rejects the null hypothesis of white noise.
C - The p-value is clearly below 0.05, so the test is conclusive.
D - The Ljung-Box test does not assess normality.
Reference: Box, Jenkins, Reinsel & Ljung (2015). Time Series Analysis, 5th Ed., Ch. 8
Page 4
2026/2027 Edition | 250 Verified Questions
ISYE 6402 Final Exam 2026-2027 QUESTIONS AND ANSWERS ALREADY GRADED A+. 100% Verified
Solutions | Updated Per Latest Guidelines | Graded A+
This comprehensive exam preparation resource contains 250 verified questions and detailed rationales
for the ISYE 6402 Time Series Analysis final exam. Each question is aligned with the latest 2026/2027
curriculum, covering all major topics from stationarity and ARIMA models to spectral analysis and
forecasting. The document is designed to ensure a thorough understanding of time series concepts and
to guarantee a passing grade with A+ performance. Every answer includes step-by-step explanations
and distractor analyses to reinforce learning.
Abstract:
This exam preparation document for ISYE 6402 Time Series Analysis provides a rigorous and comprehensive
review of all core concepts required for the final exam. The material is organized into content areas that mirror the
course syllabus, ensuring systematic coverage. Each of the 250 verified questions is accompanied by a detailed
rationale that explains the correct answer and analyzes common distractors, thereby deepening conceptual
understanding. The document emphasizes practical application through real-world data examples and forecasting
exercises. Special attention is given to model identification, estimation, and diagnostic checking, as well as
advanced topics such as spectral analysis and cointegration. By working through these questions, students will
develop the analytical skills necessary to excel in the exam and apply time series methods in research or industry.
The content has been updated to reflect the latest academic year and incorporates feedback from previous exam
cycles. This resource is designed to be the sole study aid needed to achieve a top grade.
Content Area Overview:
Content Area Questions Key Topics Weight
Stationarity and Time Series 1-40 Stationarity, autocorrelation, partial 16%
Basics autocorrelation, white noise, transformations
ARMA and ARIMA Models 41-90 AR, MA, ARMA, ARIMA, identification, 20%
estimation, diagnostic checking
Seasonal Models and SARIMA 91-130 Seasonal differencing, SARIMA, seasonal 16%
decomposition, model selection
Forecasting 131-170 Point forecasts, prediction intervals, forecast 16%
accuracy measures, rolling forecasts
Spectral Analysis 171-210 Periodogram, spectral density, filtering, 16%
frequency domain analysis
Advanced Topics 211-250 GARCH, state-space models, cointegration, 16%
multivariate time series
Page 1
,Q1. Given a time series with a unit root, which of the following transformations is most appropriate
to achieve stationarity?
A. First differencing
B. Log transformation
C. Seasonal differencing
D. Box-Cox transformation
Correct Answer: A. First differencing
Rationale: A unit root implies non-stationarity that can be removed by first differencing, which
transforms the series to its changes. Log transformation stabilizes variance but does not remove unit
roots. Seasonal differencing addresses seasonal unit roots. Box-Cox transformation is for variance
stabilization.
Why Wrong:
B - Log transformation stabilizes variance but does not remove unit roots.
C - Seasonal differencing is for seasonal unit roots, not a simple unit root.
D - Box-Cox transformation is for variance stabilization, not for removing unit roots.
Reference: Brockwell & Davis (2016). Introduction to Time Series and Forecasting, 3rd Ed., Ch. 5
Q2. In spectral analysis, the periodogram is defined as the squared magnitude of the Fourier
transform of the data. Which of the following is a key limitation of the periodogram as an estimator
of the spectral density?
A. It is biased but consistent.
B. It is unbiased but inconsistent.
C. It is both biased and inconsistent.
D. It is asymptotically unbiased and consistent.
Correct Answer: B. It is unbiased but inconsistent.
Rationale: The periodogram is an unbiased estimator of the spectral density for a given frequency, but its
variance does not decrease with sample size, making it inconsistent. Smoothing methods are needed to
achieve consistency.
Why Wrong:
A - The periodogram is unbiased, not biased.
C - The periodogram is unbiased, but it is inconsistent.
D - The periodogram is unbiased but inconsistent, not consistent.
Reference: Shumway & Stoffer (2017). Time Series Analysis and Its Applications, 4th Ed., Ch. 4
Page 2
,Q3. Consider a state space model with observation equation y_t = _t + _t and state equation _t =
_{t-1} + _t, where _t and _t are independent white noise processes with variances _^2 and _^2
respectively. This model is equivalent to which of the following?
A. ARIMA(0,1,1) model
B. ARIMA(1,0,0) model
C. ARIMA(0,0,1) model
D. ARIMA(1,1,0) model
Correct Answer: A. ARIMA(0,1,1) model
Rationale: This is the local level model, which yields an ARIMA(0,1,1) representation for y_t. The first
difference of y_t is an MA(1) process with parameter depending on the signal-to-noise ratio.
Why Wrong:
B - ARIMA(1,0,0) is an AR(1) model, which does not have a unit root.
C - ARIMA(0,0,1) is an MA(1) model for the level, not differenced.
D - ARIMA(1,1,0) is an ARI(1,1) model, not equivalent to the local level.
Reference: Durbin & Koopman (2012). Time Series Analysis by State Space Methods, 2nd Ed., Ch. 2
Q4. Two nonstationary time series, X_t and Y_t, are cointegrated if there exists a linear combination
Z_t = Y_t - X_t that is stationary. Which of the following statements about cointegration is correct?
A. Both series must be integrated of order 1, and the cointegrating vector is unique.
B. Both series must be integrated of the same order, and the cointegrating vector is unique up to scalar
multiplication.
C. Both series must be integrated of order 1, and there can be multiple cointegrating vectors if there
are more than two series.
D. Both series must be stationary, and cointegration is equivalent to correlation.
Correct Answer: B. Both series must be integrated of the same order, and the cointegrating vector is
unique up to scalar multiplication.
Rationale: For cointegration, each series must be integrated of the same order d (typically d=1). The
cointegrating vector is unique up to a scalar multiple when there are two series. With more than two
series, multiple cointegrating vectors may exist.
Why Wrong:
A - The order of integration need not be exactly 1; but it must be the same for both.
C - For two series, the cointegrating vector is unique up to scaling, not multiple.
D - Cointegration requires nonstationary series, not stationary.
Reference: Hamilton (1994). Time Series Analysis, Ch. 19
Page 3
, Q5. In a GARCH(1,1) model, the unconditional variance of the error term exists if and only if:
A. 1 + 1 < 1
B. 1 + 1 1
C. 1 + 1 > 1
D. 1 + 1 = 1
Correct Answer: A. 1 + 1 < 1
Rationale: For a GARCH(1,1) model, the unconditional variance is finite and constant if ±1 + ²1 < 1. If
1 + 1 = 1, the process is IGARCH and the unconditional variance is infinite.
Why Wrong:
B - When 1+1=1, unconditional variance is infinite (IGARCH).
C - If 1+1>1, the variance explodes over time.
D - Equality to 1 corresponds to IGARCH, not finite variance.
Reference: Tsay (2010). Analysis of Financial Time Series, 3rd Ed., Ch. 3
Q6. A researcher fits an ARMA(2,2) model to a time series and obtains residuals. The Ljung-Box
test on the residuals yields a p-value of 0.03 at lag 10. Which of the following is the most appropriate
conclusion at the 5% significance level?
A. The residuals are white noise; the model is adequate.
B. There is evidence of remaining autocorrelation in the residuals; the model may be inadequate.
C. The test is inconclusive because the p-value is close to 0.05.
D. The residuals are normally distributed.
Correct Answer: B. There is evidence of remaining autocorrelation in the residuals; the model may
be inadequate.
Rationale: A p-value of 0.03 < 0.05 indicates significant autocorrelation in the residuals at lag 10,
suggesting the model does not capture all temporal dependencies. The Ljung-Box test does not test
normality.
Why Wrong:
A - A p-value less than 0.05 rejects the null hypothesis of white noise.
C - The p-value is clearly below 0.05, so the test is conclusive.
D - The Ljung-Box test does not assess normality.
Reference: Box, Jenkins, Reinsel & Ljung (2015). Time Series Analysis, 5th Ed., Ch. 8
Page 4
