The least-squares regression line is the formal line of best fit for quantitative data, chosen by a mathematical rule rather than by eye, so predictions come from one consistent linear model.

What the least-squares regression line represents

When two quantitative variables show an approximately linear pattern, statisticians use the least-squares regression line to model that relationship. It is not just any line drawn through a scatterplot. It is the single line chosen by a rule that makes the overall prediction errors as small as possible in a squared sense.

The least-squares regression line is often called the LSRL. In AP Statistics, it is the standard regression line used for linear prediction from an explanatory variable to a response variable.

This equation describes predicted values, not necessarily the actual observed values. In most real data sets, points do not lie exactly on the line, so the LSRL balances the data instead of matching each observation perfectly.

Because the model is linear, the predicted change from one value of $x$ to another follows a straight-line pattern. The line may describe the data well or poorly, but the LSRL is still the line selected by the least-squares rule.

Why the method is called least squares

The key idea is that each observed point has some vertical distance from the regression line.

Scatterplot with the least-squares regression line and several residuals drawn as vertical line segments from each observed point to the fitted line. This directly illustrates that residuals are vertical differences $y_i-\hat{y}_i$ (in response-variable units) and motivates why the least-squares criterion focuses on vertical prediction errors. Source

That distance is measured through the residual.

A positive residual means a point lies above the line, and a negative residual means it lies below the line.

The LSRL is the line with the smallest possible value of $SSE$. Squaring matters for two important reasons:

• Negative and positive residuals do not cancel each other out.

• Larger residuals count more heavily than smaller ones, so points far from the line vertically have more effect on the fit.

Least squares is one criterion among several possible fitting rules, but it gives a definite and reproducible answer. If two people use the same data and the same variables, they will get the same least-squares regression line.

The point of means property

An important property of the LSRL is that it always passes through the point of means.

This property gives a useful check on regression output. If you know the sample means of $x$ and $y$, the least-squares regression line must go through that point. In symbols, when $x=\bar{x}$, the predicted value from the LSRL is $\hat{y}=\bar{y}$.

The line does not need to pass through any other actual data point, and in many data sets it passes through none of them exactly. The point of means shows that the fitted line is centered around the average values of the two variables, even though individual observations may be above or below it.

Estimating the regression-line parameters

The LSRL estimates the regression-line parameters, meaning it provides numerical values that determine the fitted line from the sample data. These values control the line’s location and steepness.

In the equation $\hat{y}=a+bx$, the two parameters are:

• $a$, the estimated y-intercept

• $b$, the estimated slope

These are estimates, not fixed values known ahead of time. If the data change, the estimated line usually changes as well. The parameters are chosen together, not separately, because the line must satisfy the least-squares condition as a whole.

In practice, technology is usually used to compute the LSRL because the estimates come from formulas that combine all the data values systematically. On the AP Statistics exam, you are often expected to understand what the regression line represents and what makes it least-squares, even when a calculator or software provides the equation.

What to emphasize for AP Statistics

For this topic, focus on the defining properties of the LSRL rather than treating it as just a calculator output.

• It is a linear model for two quantitative variables.

• It is chosen by minimizing the sum of squared residuals.

• The errors being minimized are vertical prediction errors in the response variable.

• The line always passes through $(\bar{x},\bar{y})$.

• Its equation uses estimated parameters taken from the sample data.

• Observed response values and predicted response values are usually different.

When reading regression output, think of the LSRL as the mathematically defined best line under one exact criterion. The line is special because its coefficients come from that minimizing rule.