2.3 Modeling with Linear Functions

Linear Functions and Modeling

Linear functions describe relationships where one variable changes at a constant rate relative to another. They're the foundation of mathematical modeling because so many real-world situations (cost analysis, revenue projection, population trends) behave linearly over a given interval. This section covers how to build linear models from data, interpret their components, and apply them to solve problems.

more resources to help you study

practice questions

Construction of Linear Models

Building a linear model means translating a real-world situation into the equation $y = mx + b$. Here's the process:

1. Identify the independent and dependent variables. The independent variable ($x$) is the input you control or that changes on its own (time, quantity produced, miles driven). The dependent variable ($y$) is the output that responds to changes in $x$ (total cost, revenue, distance remaining).

2. Determine the slope ($m$). The slope is the constant rate of change between the two variables. If you have two data points, calculate it with:

$m = \frac{y_2 - y_1}{x_2 - x_1}$

For example, if producing 10 units costs $70 and producing 20 units costs $120, then $m = \frac{120 - 70}{20 - 10} = \frac{50}{10} = 5$ dollars per unit.

1. Identify the y-intercept ($b$). This is the value of $y$ when $x = 0$. Sometimes the problem gives it directly (an initial fee, a starting balance). Other times, substitute a known point and the slope into $y = mx + b$ and solve for $b$.

Using the example above: $70 = 5(10) + b$, so $b = 20$.

1. Write the model. Combine the slope and y-intercept: $y = 5x + 20$.

Components of Linear Functions

Once you have a model, you need to interpret what each piece means in context.

Slope ($m$): Describes how the dependent variable changes per one-unit increase in the independent variable. In the cost model $y = 5x + 20$, the slope of 5 means each additional unit produced adds $5 to the total cost.

Y-intercept ($b$): Describes the starting value of the dependent variable before the independent variable has any effect. In $y = 5x + 20$, the y-intercept of 20 means there's a fixed cost of $20 even if zero units are produced (think of it as setup costs or overhead).

Domain and range should reflect the context of the problem, not just "all real numbers."

• The domain is the set of realistic input values. If $x$ represents units produced, the domain might be $x \geq 0$ (you can't produce negative units). If the model only applies up to a production capacity of 500, then $0 \leq x \leq 500$.
• The range is the set of output values that result from the domain. For $y = 5x + 20$ with $0 \leq x \leq 500$, the range is $20 \leq y \leq 2520$.

Always ask: does this value make sense in the real-world scenario?

Application of Linear Models

Evaluating the model (finding $y$ from $x$): Substitute the given $x$-value directly into the equation. If the revenue model is $y = 10x + 50$ and 100 units are sold:

$y = 10(100) + 50 = 1{,}050$

So selling 100 units generates $1,050 in revenue.

Solving for $x$ from a given $y$: Work backward through the equation. To find how many units produce $1,000 in revenue using $y = 10x + 50$:

1. $1{,}000 = 10x + 50$
2. $950 = 10x$
3. $x = 95$ units

Making predictions: You can use the model to estimate values you don't have data for. Interpolation estimates values within the range of your known data and tends to be reliable. Extrapolation estimates values beyond your data and carries more risk, since the linear trend may not hold indefinitely.

For instance, if your cost data covers 0 to 100 units, predicting the cost at 50 units is interpolation. Predicting the cost at 500 units is extrapolation, and the actual cost could differ significantly if, say, bulk discounts or new equipment costs kick in.

Limitations and assumptions are worth thinking about on any modeling problem:

• Linear models assume a constant rate of change. In reality, rates often shift (economies of scale, market saturation, diminishing returns).
• External factors like inflation, competition, or supply disruptions aren't captured by a simple linear equation.
• The model is only as good as the data it's built from. A model based on two data points is far less reliable than one based on twenty.

Data Analysis and Model Fitting

When you're given raw data rather than a neat equation, you need to determine whether a linear model is appropriate and then find the best one.

• Scatter plots let you visualize the relationship. Plot each data pair as a point and look for a roughly linear pattern.
• Correlation measures the strength and direction of the linear relationship. A correlation coefficient ($r$) close to $1$ or $-1$ indicates a strong linear association; close to $0$ suggests the relationship isn't linear.
• Linear regression (often done with a calculator or software) finds the line of best fit by minimizing the overall distance between the data points and the line.
• Residuals are the differences between observed values and the values your model predicts: $\text{residual} = y_{\text{observed}} - y_{\text{predicted}}$. If residuals are small and randomly scattered (no clear pattern), the linear model fits well. If residuals show a curve or trend, a linear model may not be the right choice.