Overview
Reading and exercises in Time Series Analysis - spring 2025
The book is available here as pdf: Time Series Analysis.
You can also buy it in the bookstore.
For each week the chapters or sections to be read are listed below.
Sample solutions are available here: Solutions. They are from the book website, but some been corrected a bit.
Week 1 - Introduction
- Book
- Chapter 1: Introduction
- Chapter 2 (not 2.8): Multivariate random variables
- Exercises
- 2.1, 2.2
- Simulation examples
- 2.3
Week 2 - Regression based methods, 1st part
The General Linear Model (GLM): Ordinary Least Squares (OLS) and Maximum Likelihood (ML) estimation.
Recursive Least Squares (RLS) and global trend models, as well as some principles needed to find a good model, such as model selection.
- Book
- 3.1: Introduction
- 3.2: The GLM, including OLS-, and ML-estimates (but we skip Section 3.2.1.4 on WLS))
- 3.3: Prediction in the GLM.
- 11.1 (with introduction): RLS.
- Additional material
- In ModellingRefence.pdf the slides about GLM and RLS (without forgetting).
- In ModelExamples.pdf the slides about the global trend models.
- Exercises
- 3.1 (maybe try solve it by hand as in the book and then check result with a computer)
- 3.4
- You can start working on Assignment 1: Section 1 to 3 is covered.
Week 3 - Regression based methods, 2nd part
- Book
- 3.2.1.4 on WLS
- 11.1.1 RLS with forgetting
- Additional material
- More on Local trend models (more to come)
- Exercises
- You can work on Assignment 1: Section 4 and 5 is about local trend models using WLS and RLS.
Week 4 - Introduction to Stochastic Processes, Operators and Linear Systems
- 4.5: Shift operators (for understanding 5.3)
- 5.1 and 5.2: Stochastic processes in general
- 5.3 (only slightly touch 5.3.2): Linear processes
- Exercises
- 5.1 (for c != 0) (Q2: see page 117 for MA(1) process)
- 5.4
- 5.7
Week 5 - AR, MA and ARMA processes
- 5.5 (disregard ‘spectra’ like (5.67), (5.72), (5.85), (5.86), (5.112)): MA, AR, and ARMA-processes
- 5.3.2: Cursory material
- 5.6: Non-stationary models
- 5.7: Optimal Prediction
- 6.4: Estimation of parameters in ARMA models
- Exercises
- Identification game
- 5.5
- 5.6 (Assume that the process is stationary and invertible.)
- 5.10 (if time allows. In Question 2 you are expected to find a recursion for gamma(k) for k>2. Skip Question 3.),
Week 6 - ACF and PACF with a focus on model order selection
Identification of univariate time series models, 1st part:
- 6.1 (with intro): Introduction
- 6.2.1 (and the introduction to 6.2 (Sec. 6.2.1 (a))): Estimation of auto-covariance and -correlation
- 6.3 (not 6.3.3): Using the SACF and SPACF for model order selection
- 6.5: Model order selection
- 6.6: Model validation
- Exercises
- ARIMA model identification
- 6.1
- 6.6
- Carry on with Assignment 2
Week 7 - Linear systems
- 4.1 (with intro): Linear Systems
- 4.4: You should disregard Theorem 4.10 and the following example. Furthermore, we shall not discuss Theorem 4.12 until we start to look at the multivariate time series.
- 6.2.2: Cross-correlation functions
- Chapter 8: Linear systems and stochastic processes
- Exercises
- R arima with and without external regressor
- 4.1 (Q. 1-2)
- 8.2
- 6.6
Week 8 - Multivariate time series
- Chapter 9 (read it coarsely, the bi-variate ARMA is good to think about, and then the rest: think like it’s the same as for uni-variate, just extended to be multi-variable in VARMAX models, as in VectorARMAX).
- MARIMA: Spliids method for parameter estimation in multi-variable ARMAX models.
- Have a look at the original paper (don’t go too much into details, it’s too cumbersome): Marima_paper
-
Have a look at the marima R package vignette (again, don’t go into details): Marima_vignette
- Exercises
- Bivariate-output ARMAX exercise: exercise_bivariate_waterheating
- If time allows, take a look at 9.1 and its solutions, just to get some points about the theoretical ACF in a bivariate model.
Week 9 - TBD
TBD
Exercises TBD
Week 10 - State space models 1st part
- 10.1: The Linear Stochastic State Space Model, Sec. 10.1
- 10.3: The Kalman filter
- Exercises
- 10.1
- Kalman filter
Week 11 - State space models 2nd part
- 10.4 (not 10.4.1): ARMA-models on state space form
- 10.6: ML-estimates of state space models
- Exercises
- Follow up on ‘Kalman Filter’ exercise from last week (estimate the gravity in the “throw” model)
- Estimation in state-space model
- 10.2
- 10.3
Week 12 - Recursive and adaptive estimation
- Chapter 11: Recursive and adaptive estimation
Execises:
- 10.4
- Get help on Assignment4
Week 13 - Final lecture
During the final lecture we will consider a number of real world problems and discuss what tools from the course can be used to solve them.
- Exercises
- Get help on Assignment4