Time Series Analysis — Questions and Answers

Question 1: What causes a forecast to be 'biased'?

Non-normality of residuals
Systematic over- or under-prediction where the mean error is consistently non-zero (Correct answer)
High variance in forecast errors
Large absolute errors

Correct answer: Systematic over- or under-prediction where the mean error is consistently non-zero

Bias occurs when a model consistently forecasts too high or too low, resulting in a non-zero mean error (ME) across forecasts.

Question 2: What is the purpose of differencing in an ARIMA model?

To achieve stationarity (Correct answer)
To remove seasonality only
To increase autocorrelation
To normalize residuals

Correct answer: To achieve stationarity

Differencing removes trends and unit roots, transforming a non-stationary series into a stationary one.

Question 3: For a SARIMA model ARIMA(1,1,1)(1,1,1)[12], what does the [12] represent?

12 AR lags
12 MA lags
Seasonal period of 12 months (Correct answer)
12 differencing steps

Correct answer: Seasonal period of 12 months

The [12] denotes the seasonal period, meaning the seasonal pattern repeats every 12 time steps (monthly data with annual seasonality).

Question 4: Over-differencing a time series typically introduces which problem?

Trend reappearance
Heteroscedasticity
Negative autocorrelation at lag 1 (Correct answer)
Perfect stationarity

Correct answer: Negative autocorrelation at lag 1

Over-differencing adds unnecessary variance and often creates a strong negative autocorrelation at the first lag.

Question 5: What is ARCH (Autoregressive Conditional Heteroscedasticity) in time series modeling?

A heteroscedastic extension of the MA model
A model where the variance of the error term depends on past squared errors (Correct answer)
A model for series with conditional mean changes
An autoregressive model with changing coefficients

Correct answer: A model where the variance of the error term depends on past squared errors

ARCH models the time-varying variance by making it a function of past squared residuals, capturing volatility clustering in financial data.

Question 6: What is a Transfer Function model?

A changepoint detection model
A model that transfers errors between series
An ARIMA model extended with lagged values of one or more input (predictor) variables (Correct answer)
A Fourier-based frequency domain model

Correct answer: An ARIMA model extended with lagged values of one or more input (predictor) variables

Transfer function (or dynamic regression) models add filtered versions of external predictors to a noise model, capturing both input-output dynamics and autocorrelated errors.

Question 7: How many cycles does business get?

Four stages (Correct answer)
Three stages
One stage
Two stages

Correct answer: Four stages

The business cycle typically consists of four distinct stages: Expansion (or Recovery/Prosperity), Peak (or Boom), Contraction (or Recession), and Trough (or Depression). These stages represent the cyclical fluctuations in economic activity over time. Understanding these four stages is fundamental to macroeconomic analysis and forecasting.

Question 8: In an ARIMA(p,d,q) model, what does the parameter 'p' represent?

Order of the autoregressive component (Correct answer)
Number of differences
Seasonal period
Order of the moving average component

Correct answer: Order of the autoregressive component

The parameter 'p' specifies the order of the autoregressive (AR) component, indicating how many lagged values are used.

Question 9: Moving averages are used to calculate the secular trend:

None of the above
Measures the seasonal variations
Smooth out the time series (Correct answer)
Give the trend in a straight line

Correct answer: Smooth out the time series

Moving averages are primarily used in time series analysis to smooth out short-term fluctuations and irregular variations in the data. By averaging values over a specific period, they help to reveal the underlying long-term trend (secular trend) more clearly, making it easier to identify the general direction of the series. This process filters out noise to highlight the main pattern.

Question 10: Which information criterion is commonly used for ARIMA model selection?

Durbin-Watson statistic
R-squared
AIC (Akaike Information Criterion) (Correct answer)
Mean Absolute Error

Correct answer: AIC (Akaike Information Criterion)

AIC balances model fit and complexity, penalizing models with more parameters to avoid overfitting.

Question 11: What is 'cointegration' in the context of multiple non-stationary time series?

Two non-stationary series share a long-run equilibrium relationship (Correct answer)
Both series are stationary
The series have identical variances
Differencing one series makes the other stationary

Correct answer: Two non-stationary series share a long-run equilibrium relationship

Cointegrated series are individually I(1) but their linear combination is stationary, implying a stable long-run relationship.

Question 12: What is an 'ensemble forecast' in time series analysis?

A forecast using only one model with multiple parameters
A forecast produced by combining predictions from multiple models or simulations (Correct answer)
A bootstrap resampling technique
A seasonal adjustment method

Correct answer: A forecast produced by combining predictions from multiple models or simulations

Ensemble forecasts aggregate outputs from multiple models (or model runs), typically outperforming any individual model by reducing variance.

Question 13: What does a significant Ljung-Box test on ARIMA residuals indicate?

The model fits perfectly
Stationarity is confirmed
Seasonality is absent
Residual autocorrelation remains — model is inadequate (Correct answer)

Correct answer: Residual autocorrelation remains — model is inadequate

A significant Ljung-Box p-value means residuals are not white noise, suggesting the ARIMA model has not captured all structure.

Question 14: Which of the following best describes the theoretical ACF of a pure MA(q) process?

It equals 1 at all lags
It cuts off to zero after lag q (Correct answer)
It decays exponentially without cutting off
It oscillates between positive and negative values indefinitely

Correct answer: It cuts off to zero after lag q

A pure MA(q) process has autocorrelations that are exactly zero for all lags beyond q, causing a sharp cutoff.

Question 15: The approximate 95% confidence bounds on an ACF plot for a series of length n are typically drawn at:

±1/n
±1/√n
±2/√n (Correct answer)
±2/n

Correct answer: ±2/√n

The approximate 95% confidence bounds for autocorrelations under the null of white noise are ±1.96/√n ≈ ±2/√n.

Question 16: An ARIMA(0,1,0) model is equivalent to which simple forecasting method?

Random walk (Correct answer)
Exponential smoothing
Simple moving average
Linear regression

Correct answer: Random walk

ARIMA(0,1,0) applies one difference with no AR or MA terms, which is the definition of a random walk model.

Question 17: Which of the following is not a time series model example?

Naive approach
Moving average
Exponential smoothing
None of the above (Correct answer)

Correct answer: None of the above

Moving average, Naive approach, and Exponential smoothing are all common and valid examples of time series forecasting models. The Moving Average model smooths out fluctuations, the Naive approach uses the last observed value as the forecast, and Exponential Smoothing assigns exponentially decreasing weights to older observations. Therefore, "None of the above" is the correct answer as all listed options are indeed time series models.

Question 18: What is the consequence of underdifferencing a time series in ARIMA modeling?

Seasonal patterns are eliminated
The series becomes over-smoothed
The model retains non-stationarity, leading to unreliable forecasts (Correct answer)
The PACF shows no significant lags

Correct answer: The model retains non-stationarity, leading to unreliable forecasts

Failing to difference enough leaves unit roots in the model, causing spurious forecasts and unreliable statistical inference.

Question 19: Why is first differencing insufficient when a series has both trend and seasonality?

Differencing amplifies seasonality
Trend differencing removes the trend but leaves seasonality, requiring additional seasonal differencing (Correct answer)
Seasonal series cannot be differenced
First differencing always produces white noise

Correct answer: Trend differencing removes the trend but leaves seasonality, requiring additional seasonal differencing

First differencing handles the trend component but seasonal patterns at period s require a separate seasonal difference to be removed.

Question 20: Which Python class in statsmodels implements the full Holt-Winters exponential smoothing model?

HoltWinters
SimpleExpSmoothing
ExponentialSmoothing (Correct answer)
ets()

Correct answer: ExponentialSmoothing

statsmodels.tsa.holtwinters.ExponentialSmoothing provides the full ETS model including additive and multiplicative trend and seasonality options.

Question 21: What is the purpose of an Intervention Analysis in time series?

To test for structural breaks automatically
To add exogenous variables to an ARIMA model for forecasting
To remove outliers from the series
To model the effect of known external events (interventions) on a time series (Correct answer)

Correct answer: To model the effect of known external events (interventions) on a time series

Intervention analysis explicitly models the impact of known events (policy changes, promotions, disasters) using indicator variables within the ARIMA framework.

Question 22: In simple exponential smoothing, what does a smoothing parameter α close to 1 indicate?

Past observations are weighted heavily
The forecast is very stable and slow to update
The forecast is identical to the mean
The forecast reacts quickly to recent changes (Correct answer)

Correct answer: The forecast reacts quickly to recent changes

An α near 1 means almost all weight is on the most recent observation, making the forecast highly responsive to the latest data point.

Question 23: In time series forecasting, what is the 'horizon' h?

The number of steps ahead being forecast (Correct answer)
The length of the training data
The number of model parameters
The seasonal period of the series

Correct answer: The number of steps ahead being forecast

The forecast horizon h specifies how many time periods into the future the forecast extends (e.g., h=12 for 12-step-ahead monthly forecasts).

Question 24: Why is splitting time series data for train/test using random sampling incorrect?

It prevents parameter estimation
It causes data leakage — future observations in training allow the model to 'see ahead' (Correct answer)
It increases variance in test set performance
It reduces training data size unnecessarily

Correct answer: It causes data leakage — future observations in training allow the model to 'see ahead'

Time series data has temporal dependencies; random splitting breaks this order, allowing future information to leak into training, producing unrealistically optimistic results.

Question 25: What is a unit root in a time series?

A constant mean value
A characteristic root equal to 1, indicating non-stationarity (Correct answer)
A root that stabilizes variance
A seasonal component with period 1

Correct answer: A characteristic root equal to 1, indicating non-stationarity

A unit root means the characteristic equation has a solution of exactly 1, which causes the series to be non-stationary with a stochastic trend.

Question 26: For a white noise process, what is the expected value of all autocorrelations at lags greater than zero?

-1
0.5
1
0 (Correct answer)

Correct answer: 0

White noise is uncorrelated across time, so all autocorrelations at lags > 0 have an expected value of zero.

Question 27: What is the Box-Jenkins methodology primarily used for?

Systematic identification, estimation, and diagnostic checking of ARIMA models (Correct answer)
Computing seasonal indices
Only visualizing time series data
Calculating moving averages

Correct answer: Systematic identification, estimation, and diagnostic checking of ARIMA models

Box-Jenkins is a three-step iterative process: identify model orders, estimate parameters, then check residual diagnostics.

Question 28: What is Granger causality in time series analysis?

A measure of cross-correlation between two series
A cointegration test
A test whether past values of series X improve predictions of series Y beyond Y's own past (Correct answer)
A test for unit roots in bivariate systems

Correct answer: A test whether past values of series X improve predictions of series Y beyond Y's own past

Granger causality tests whether lagged values of X add statistically significant predictive power for Y after controlling for Y's own lags.

Question 29: What is 'dynamic time warping' (DTW) used for in time series analysis?

Differencing two series to find cointegration
Aligning seasonal indices across years
Measuring similarity between two time series that may be shifted or distorted in time (Correct answer)
Computing cross-correlations at multiple lags

Correct answer: Measuring similarity between two time series that may be shifted or distorted in time

DTW finds the optimal alignment between two sequences by allowing flexible time shifts, making it a robust distance measure for time series classification and clustering.

Question 30: In ARIMA modeling, what does 'd=2' imply?

The series was differenced twice to achieve stationarity (Correct answer)
The seasonal period is 2
There are 2 MA terms
There are 2 AR terms

Correct answer: The series was differenced twice to achieve stationarity

d=2 means second-order differencing was applied, removing both linear and quadratic trends from the series.

Time Series Analysis

A comprehensive assessment covering statistical methods for analyzing time-ordered data, including ARIMA modeling, forecasting techniques, and frequency-domain spectral analysis. Commonly tested in graduate statistics, data science, and econometrics programs.

Questions

35 min

Time Limit

70%

To Pass

Checking for saved progress...

Exam Rules

You can skip questions and return to them later
Flag questions for review before submitting
No feedback shown until you submit the entire exam
Unanswered questions count as wrong — answer everything
10 pretest questions are mixed in and don't affect your score
Timer auto-submits when time runs out
Your progress is auto-saved every 30 seconds