Econometric Theory/Ordinary Least Squares (OLS)

< Econometric Theory

Ordinary Least Squares or OLS is one of the simplest (if you can call it so) methods of linear regression. The goal of OLS is to closely "fit" a function with the data. It does so by minimizing the sum of squared errors from the data.

Why we Square Errors before SummingEdit

We are not trying to minimize the sum of absolute errors, but rather the sum of squared errors. Let's take a brief look at our sweater story again.

Model A
Model B

model data point error from line
A 1 5
A 2 10
A 3 -5
A 4 -10
B 1 3
B 2 -3
B 3 3
B 4 -3

Notice that the Sum of Model A is 5 + 10 - 5 -10 = 0 and that the Sum of Model B is  3 - 3 + 3 - 3 = 0

Both Models sum to 0 and both are great fits! NO!!

So to account for the signs, whenever we sum errors, we square the terms first.

The ModelEdit

These two models each have an intercept term \alpha, and a slope term \beta (some textbooks use \beta_0 instead of \alpha and \beta_1 instead of \beta, this is a much better approach once we move to multivariate formulas). We can represent an arbitrary single variable model with the formula: y_i = \alpha + \beta x_i + u_i The y-values are related to the x-values given this formula. We use the subscript i to denote an observation. So y_1 is paired with x_1, y_2 with x_2, etc. The u_t term is the error term, which is the difference between the effect of x_i and the observed value of y_i.

Unfortunately, we don't know the values of \alpha, \beta or u_t. We have to approximate them. We can do this by using the ordinary least squares method. The term "least squares" means that we are trying to minimize the sum of squares, or more specifically we are trying to minimize the squared error terms. Since there are two variables that we need to minimize with respect to (\alpha and \beta), we have two equations:
f = \Sigma u_i^2 = \Sigma (y_i - \alpha - \beta x_i)^2
\frac{\partial f}{\partial \alpha} = -2\Sigma (y_i - \alpha - \beta x_i) = 0
\frac{\partial f}{\partial \beta} = -2\Sigma (y_i - \alpha - \beta x_i) x_i = 0
Call the solutions to these equations \hat \alpha and \hat \beta. Solving we get:
\hat \alpha = \bar y - \hat \beta \bar x
\hat \beta = \frac{\Sigma(x_i - \bar x)y_i}{\Sigma(x_i - \bar x)^2}
Where \bar y = \frac{\Sigma y_i}{n} and \bar x = \frac{\Sigma x_i}{n}. Computing these results can be left as an exercise.

It is important to know that \hat \alpha and \hat \beta are not the same as \alpha and \beta because they are based on a single sample rather than the entire population. If you took a different sample, you would get different values for \hat \alpha and \hat \beta. Let's call \hat \alpha and \hat \beta the OLS estimators of \alpha and \beta. One of the main goals of econometrics is to analyze the quality of these estimators and see under what conditions these are good estimators and under which conditions they are not.

Once we have \hat \alpha and \hat \beta, we can construct two more variables. The first is the fitted values, or estimates of y:
\hat y_i = \hat \alpha + \hat \beta
The second is the estimates of the error terms, which we will call the residuals:
\hat u_i = y_i - \hat y_i
These two variables will be important later on.