M8 - Model Diagnostics, Selection and Performance

Learning Objectives

The learning goals for this module are:

• Discuss model selection criteria: Akaike and Bayesian Information Criteria;
  
• Discuss residual analyis;
  
• Introduce common forecast performance/accuracy metrics;
  
• Learn how to compute forecast accuracy in R.
  

Slides

Here is a link to the slide deck used in class.

Resources

• Time Series Analysis with Applications - Cryer and Shan - Chapter 8: Diagnostics
  

Recordings

• Model selection and performance
  

• Resiual analysis
  

Optional Readings

If you want to learn more about parameter estimation for the ARIMA model, please refer to the additional material below. The slides will go over how to estimate the autoregressive coefficient (i.e. PACF values), moving average coefficent and variance of residuals.

• ARIMA Parameter Estimation

• Time Series Analysis with Applications - Cryer and Shan - Chapter 7: Parameter Estimation
  

Deliverables

There is no assignment associated with this module but you will have a chance to explore this content when checking accuracy of the model you develop for the load forecasting cometition and the final project. Visit the Assignments tab to check deadlines for A08 and the Final Project.