At the beginning of the chapter, we discussed how the samples are distributed after the linear regression model has been fitted:
Clearly, the Gaussian itself is agnostic to the way the coefficients have been determined, and by employing a standard method such as OLS or the closed-form expression, we are implicitly relying only on the dataset. Our assumption is that we have enough samples to represent the underlying data generating process correctly and the coefficients must be chosen in a way that minimizes the squared error. However, we may have some prior beliefs about the distribution of all parameters (for example, we could imagine that θi is drawn from a Gaussian distribution) and we would like to include this piece of information in our model. As we are going to discuss in Chapter 6, Naive Bayes and Discriminant Analysis (for further details, please...
