Vellore – 632014, Tamil Nadu, India
SCHOOL OF COMPUTER SCIENCE AND ENGINEERING
Fall Semester 2025-2026
BCSE336L- FDA
Module 2 Question Bank
1. Define stationarity in time series analysis. Explain why stationarity is crucial for AR,
MA, and ARMA models, and describe two methods to achieve it.
2. Compare AR(p) and MA(q) models in terms of:
Model structure.
ACF/PACF behavior.
Use cases in finance.
3. How would you identify an ARMA(1,1) model using ACF and PACF plots? Contrast
its patterns with pure AR(1) and MA(1) models.
4. Explain the Augmented Dickey-Fuller (ADF) test for unit roots. Interpret the following
ADF test output:
Test statistic: −2.8
Critical value (5%): −2.86
5. Relate Simple Exponential Smoothing (SES) to an ARIMA model. Write the SES
forecasting equation and explain the role of the smoothing parameter α.
6. Define invertibility in MA models. Why is it important, and how can you check it for
an MA(1) model with θ1=0.8?
7. Compare AIC and BIC for time series model selection. Which one favors simpler
models, and why?
8. Differentiate a random walk from a trend-stationary series. Provide an example of each
in finance.
9. Describe the Ljung-Box test’s purpose in residual analysis. Interpret a p-value of 0.03
for lag 10.
10. What is the Extended Autocorrelation Function (EACF), and how does it help identify
ARMA(p,q) orders?
11. Given the following time series data for a stock's log returns:
x=[0.02,−0.01,0.03,0.015,−0.005]
Estimate the AR(1) model parameter ϕ1 using least squares.
Check if the model is stationary.
12. An MA(1) model is fitted to stock returns with θ1=−0.4 and the last two residuals ϵt−1
=0.1, ϵt=−0.2.
Calculate the 1-step-ahead forecast xt+1.
Compute the 2-step-ahead forecast xt+2.
13. The ACF and PACF plots of a financial time series show:
ACF: Significant at lag 1, then cuts off.
PACF: Decays exponentially.
Identify the appropriate model (AR/MA/ARMA).
Justify your choice.
Vellore – 632014, Tamil Nadu, India
SCHOOL OF COMPUTER SCIENCE AND ENGINEERING
Fall Semester 2025-2026
BCSE336L- FDA
Module 2 Question Bank
14. The ADF test statistic for a stock price series is −1.9, and the 5% critical value is −2.89.
Interpret the result.
What transformation would you apply?
15. Given smoothing parameter α=0.3 and the last observed value yt=50, with the last
forecast ŷt=48:
Compute the next forecast yt+1.
Explain how α affects responsiveness.
16. An ARMA(1,1) model has ϕ1=0.6, θ1=−0.3, and ϵt∼ N (0,1).
Write the model equation.
Simulate one step given xt−1=1.2 and ϵt−1=−0.5.
17. Two stocks have log returns:
Stock A: [0.01,0.02,−0.01]
Stock B: [−0.005,0.015,0.01]
Compute the covariance matrix.
Interpret the off-diagonal element.
18. An AR(1) model has AIC = -120.5, and an MA(1) model has AIC = -118.3.
Which model is better?
Why might you still choose the MA(1) model?
19. For a fitted ARMA(1,1) model, the Ljung-Box test yields Q(5)=8.4 (p-value = 0.08).
Interpret the result. Is the model adequate?