Exam SRM Details

The Linear Models section is the largest portion of the exam. We will introduce many models that fall under this umbrella.

We will begin with a fundamental statistical learning method known as simple linear regression and then extend it to multiple linear regression. Along the way, we will examine the assumptions underlying these models. We will then introduce methods that make adjustments to this framework in many different ways.

Without going into too much detail in this short summary, the main idea of this section is to start small and gradually build on what we have learned.

In this section, we discuss tree-based methods. These methods are based on decision trees. A decision tree partitions the predictor space into multiple regions.

Decision trees can be used to predict either a numerical target variable or a categorical target variable. We begin with regression trees and then move to classification trees. After that, in the next section, we discuss ensemble methods, which combine many trees within a single model.

We now pivot to methods used to study data without a specified response variable. This collection of methods is known as unsupervised learning. Without a response variable, the goal is no longer prediction, but rather to uncover hidden structure and gain insight into the underlying patterns in the data. In this course, we will cover two unsupervised learning methods: principal components analysis (PCA) and clustering.

Now that we have covered many statistical learning methods, we turn to several overarching concepts that apply across all of them. Some of these ideas may have been introduced earlier, but the goal of this section is to further clarify them and bring them together in one place.

We will also revisit resampling methods and examine their role in measuring model accuracy.

Welcome to Section 6, the final section of the course, where we venture into time series analysis. A time series consists of observations indexed by time, typically recorded at regular intervals.

Time series analysis provides statistical tools for modeling and understanding data that evolve over time. This is especially important for time-dependent phenomena such as stock prices, economic indicators, weather patterns, and traffic conditions.