Linear functions

A linear function is a function of the form

$f(x) = mx + b$		Function form
$y = mx + b$		Equation form

with $m$ and $b$ being fixed numbers (the names "$m$" and "$b$" are traditional)

Role of m

Numerically:
• A change of 1 unit in $x$ results in a change of $m$ units in $y.$
• A change of $2$ units in $x$ results in a change of $m \times 2$ units in $y.$
• ...
• In general, a change of $\Delta x$ units in $x$ results in a change of $\Delta y = m\Delta x$ units in $y,$ so that
  $m = \frac{\text{Change in }y}{\text{Change in }x} = \frac{\Delta y}{\Delta x}$
  Graphically:
• $m =$ Slope $= \frac{\text{Rise}}{\text{Run}} = \frac{\Delta y}{\Delta x}$

Role of b

Numerically:
• When $x = 0,\ \ y = b$ (Equation form), or $f(0) = b.$ (Function form)
Graphically:
• $b =$ $y$-intercept

$\color{blue}{b}$ For these reasons, the equation $y = mx + b$ is often called the slope-intercept form of the equation of a line.

How to draw the graph of a linear function

There are two good ways to draw the graph of a linear function:

1. Write the function in $y = mx+b$ form, and draw the line with $y$-intercept $b$ and slope $m$.
2. Find the $x$- and $y$-intercepts and draw the line going through those two points. To find the $x$-intercept of a line, set $y = 0$ in its equation and solve for $x$. To find the $y$-intercept, set $x = 0$ and solve for $y$. This method works only if the line does not pass through the origin. If it does, then you will need to plot an extra point or use the first method.

Finding a linear equation from data: How to make a linear model

Point and slope

An equation of the line through the point $(x_1, y_1)$ with slope $m$ is given by

$y = mx + b$

where

$b = y_1 - mx_1$.

When to use this method

1. You can use this method whenever you know the slope and the coordinates of a point on the line. The formula does not apply if the slope is undefined.
2. If you already know the slope $m$ and $y$-intercept $b$, then you can just write down the linear function directly as
$y = mx + b$
by just substituting the values of $m$ and $b$.

Horizontal and vertical lines

An equation of the horizontal line through $(x_1, y_1)$ is

$y = y_1$

An equation of the vertical line through $(x_1, y_1)$ is

$x = x_1$

Interpretation of the slope in applications

The slope of the line $y = mx + b$ is the rate at which $y$ is changing per 1-unit of change in $x$. The units of measurement of the slope are units of $y$ per unit of $x$.

If $y$ is displacement (position along a line) and $x$ is time, then the slope represents velocity. Its units are units of displacement per unit time (for example, meters per second)

If $y$ is cost and $x$ is the number of items purchased, then the slope represents marginal cost. Its units are units of cost per item (for example, dollars per item).

Regression: Residuals and sum of squares error

Linear regression is a method of finding the linear equation that comes closest to fitting a collection of data points.

Observed and predicted values

Suppose we are given a collection of data points $(x_1, y_1), ..., (x_n, y_n)$. The $n$ quantities $y_1, y_2, ..., y_n$ are called the observed values of $y$.

If we approximate the data by the linear equation $\hat{y} = mx + b$, then the predicted values are

$\hat{y}_1 = m(x_1) + b$,
$\hat{y}_1 = m(x_2) + b$,
...,
$\hat{y}_n = m(x_n) + b.$

Residuals and sum of squares error

If we model a collection of data points $(x_1, y_1), ..., (x_n, y_n)$ with a linear equation as above, then the residuals are the $n$ quantities (Observed Value − Predicted Value):

$(y_1 - \hat{y}_1), \ (y_2 - \hat{y}_2), \ \dots (y_n - \hat{y}_n)$

↑ = Residual

The sum of squares error (SSE) is the sum of the squares of the residuals:

SSE $= (y_1 - \hat{y}_1)^2 + (y_2 - \hat{y}_2)^2 + \cdots + (y_n - \hat{y}_n)^2$

Regression: Finding the regression line

The regression line (least squares line, best-fit line) associated with the points $(x_1, y_1), ..., (x_n, y_n)$ is the line that gives the minimum value for SSE.

The regression line is given by

$y = mx + b$,

where

$m = \frac{n\left(\sum xy\right) - \left(\sum x\right)\left(\sum y\right)}{n\left(\sum x^2\right) - \left(\sum x\right)^2},$
$b = \frac{\left(\sum y\right) - m\left(\sum x\right)}{n}$.

Here, $\sum $ stands for "the sum of." Thus, for example,

$\sum x =$ Sum of the $x$-values $= (x_1 + x_2 + \dots + x_n)$
$\sum xy =$ Sum of the products $xy$ $= (x_1y_1 + x_2y_2 + \dots + x_ny_n)$
$\sum x^2 =$ Sum of the squares of the $x$-values $= (x_1^2 + x_2^2 + \dots + x_n^2)$

On the other hand,

$\left(\sum x\right)^2 =$ Square of $\sum x \quad = \quad $ Square of the sum of the $x$-values.

Finally,

$n =$ Number of data points.