Choosing Model Types — I can choose an appropriate model type (linear/exponential/none) for a data set and justify. - Grade 9 Math

SECTION 1

Historical Context & Motivation

For centuries, scientists and mathematicians have searched for patterns in the world around them. Whether tracking the growth of a population, predicting the path of a projectile, or forecasting the balance in a savings account, the central challenge has always been the same: what kind of equation best describes the data? Choosing the wrong model can lead to wildly inaccurate predictions, while selecting the right one can unlock powerful insights. The ability to distinguish between linear growth, exponential growth, and data that follows neither pattern is one of the most practical skills in modern mathematics.

~300 BCE • Euclid & Proportional Reasoning

Ancient Greek mathematicians studied proportional relationships — the idea that two quantities can increase at a constant rate relative to each other. This laid the groundwork for what we now call linear models.

1798 • Malthus & Population Growth

Thomas Malthus published his essay on population, arguing that human populations grow exponentially while food production grows only linearly. This was one of the first prominent uses of comparing linear and exponential models to make predictions about the real world.

1800s • Least-Squares Regression

Carl Friedrich Gauss and Adrien-Marie Legendre developed methods to fit the best possible line or curve to a set of data points. These techniques became the foundation of modern data modeling and statistics.

2000s • Big Data & Model Selection

With the explosion of data in the digital age, choosing the correct model type has become essential in fields from medicine to finance. Algorithms now automate model selection, but understanding the fundamentals remains critical for interpreting results.

The core question this lesson addresses is straightforward but powerful: given a table of values, a scatter plot, or a real-world scenario, how do you determine whether a linear model, an exponential model, or neither is the best fit? And just as importantly, how do you justify your choice with mathematical evidence?

SECTION 2

Core Principles & Definitions

Before you can choose a model, you need to understand what makes each type tick. A mathematical model is simply an equation or function that approximates a real-world relationship. The two most common types you will encounter in Grade 9 are linear and exponential models, but sometimes neither of those is a good fit. Recognizing the signature patterns of each type is the key to making the right choice.

Linear Model

Data changes by a constant amount (called the slope) for each equal step in the input. The graph is a straight line. Example: earning $12 per hour — every hour adds the same amount.

Exponential Model

Data changes by a constant ratio (called the growth or decay factor) for each equal step in the input. The graph is a curve that accelerates or decelerates. Example: bacteria doubling every hour.

Neither (No Good Fit)

The data does not show a constant difference or a constant ratio. It might follow a quadratic, periodic, or other pattern — or it may be random. Forcing a linear or exponential model onto this data would produce unreliable predictions.

First Differences

The differences between consecutive y-values when x increases by equal steps. If first differences are roughly constant, the data is linear.

Common Ratio

The ratio of consecutive y-values when x increases by equal steps. If this ratio is roughly constant, the data is exponential.

✦

KEY TAKEAWAY

Think of it like a road trip. A linear model is like a car driving at a perfectly steady speed — you cover the same distance every hour. An exponential model is like a car that doubles its speed every hour — at first it feels normal, but soon it's moving incredibly fast. And sometimes the car is swerving, stopping, and speeding up randomly — that's when neither model works.

SECTION 3

Visual Explanation — Recognizing the Shapes

One of the fastest ways to identify a model type is to look at the shape of the data when it is plotted on a coordinate plane. The diagram below shows the three scenarios you will encounter. Notice how the linear data forms a straight line, the exponential data curves upward (or downward for decay), and the "neither" data follows an irregular pattern that cannot be captured by a single straight line or a smooth exponential curve.

The linear graph on the left shows points that fall along a straight line with equal spacing vertically. The exponential graph in the center shows a curve that starts slow and then rises sharply. The neither graph on the right shows scattered data with no consistent pattern.

Looking at a graph is a great first step, but it is not always enough. Sometimes data that looks slightly curved might actually be linear with some measurement error, and data that appears straight over a short range might actually be part of an exponential curve. That is why we also use numerical tests — first differences and common ratios — which we will explore in the next section.

SECTION 4

Mathematical Framework

Each model type has a general equation and a specific numerical test you can apply to a data table. Here are the formulas you need to know, along with the diagnostic tests for identifying each type.

LINEAR MODEL

y = mx + b

m = slope (constant rate of change); b = y-intercept (starting value when x = 0). The test: compute first differences (y₂ − y₁, y₃ − y₂, …). If they are approximately equal, the data is linear.

EXPONENTIAL MODEL

y = a · bˣ

a = initial value (y-value when x = 0); b = growth factor (b > 1 for growth, 0 < b < 1 for decay). The test: compute common ratios (y₂ ÷ y₁, y₃ ÷ y₂, …). If they are approximately equal, the data is exponential.

FIRST DIFFERENCES TEST

Δy = yₙ₊₁ − yₙ

Calculate the difference between each pair of consecutive y-values. If all the first differences are roughly the same number, the relationship is linear with slope equal to that constant difference (assuming equal x-steps).

COMMON RATIO TEST

r = yₙ₊₁ ÷ yₙ

Calculate the ratio of each consecutive pair of y-values. If all the ratios are roughly the same number, the relationship is exponential with growth factor equal to that constant ratio (assuming equal x-steps).

⚠

Important Note

These tests work reliably only when the x-values are equally spaced (e.g., x = 1, 2, 3, 4 or x = 0, 5, 10, 15). If the x-values are unequal, you need to adjust your calculations before comparing differences or ratios.

SECTION 5

The Decision Flowchart

When you are given a data set and asked to choose a model, it helps to follow a systematic decision process. The flowchart below summarizes the steps. Start by checking whether the x-values are equally spaced, then compute first differences and common ratios, and finally make your classification based on which test yields approximately constant results.

Follow this flowchart every time you need to classify a data set. Start at the top, compute both tests, then follow the branch that matches. Always end by justifying your choice using the numerical evidence you computed.

A strong justification includes specific numbers. For example, saying "the first differences are all approximately 5, so the data is linear" is much stronger than simply saying "it looks like a line." On an exam, referencing your computed differences or ratios is what earns full credit.

SECTION 6

Worked Example

Let's walk through a complete example. Suppose you are given the following data table and asked to determine whether the relationship is linear, exponential, or neither.

x	y
0	3
1	6
2	12
3	24
4	48

Data set for the worked example

Classifying the Data Set

1
Step 1 — Check x-spacing

The x-values go 0, 1, 2, 3, 4 — each increases by 1. Since the x-values are equally spaced, we can proceed directly to computing first differences and common ratios.

x-values are equally spaced (step = 1) ✓
2
Step 2 — Compute first differences

Subtract each y-value from the next: 6 − 3 = 3, 12 − 6 = 6, 24 − 12 = 12, 48 − 24 = 24. The first differences are 3, 6, 12, 24. These are clearly not constant — they keep changing.

First differences: 3, 6, 12, 24 — NOT constant → not linear
3
Step 3 — Compute common ratios

Divide each y-value by the previous one: 6 ÷ 3 = 2, 12 ÷ 6 = 2, 24 ÷ 12 = 2, 48 ÷ 24 = 2. The common ratios are all exactly 2. This is constant!

Common ratios: 2, 2, 2, 2 — CONSTANT → exponential
4
Step 4 — Identify the equation

Since the data is exponential, we use y = a · bˣ. The initial value (when x = 0) is a = 3, and the common ratio is b = 2. So the equation is y = 3 · 2ˣ.

Model: y = 3 · 2ˣ (exponential growth)
5
Step 5 — Write the justification

An exponential model is appropriate because the common ratios between consecutive y-values are all equal to 2. This means the y-value doubles for every unit increase in x, which is the defining characteristic of exponential growth. The first differences (3, 6, 12, 24) are not constant, confirming that a linear model would be inappropriate.

Justified: exponential model with constant ratio of 2

SECTION 7

Comparing Linear vs. Exponential

Understanding the differences between linear and exponential models helps you make faster, more confident decisions. The table below provides a side-by-side comparison of their most important features.

Feature	Linear Model	Exponential Model
Equation	y = mx + b	y = a · bˣ
Type of change	Constant addition (or subtraction)	Constant multiplication (or division)
Graph shape	Straight line	Curve (J-shape for growth, decreasing for decay)
Key numerical test	First differences are constant	Common ratios are constant
Real-world example	Saving $50 per week	A viral post doubling in shares every day
Long-term behavior	Grows steadily without bound	Grows explosively fast (or decays toward zero)

✦

KEY TAKEAWAY

Here's a quick way to remember the difference: if you add the same amount each time, it's linear; if you multiply by the same amount each time, it's exponential. Think of your allowance — getting a flat $10 raise every year is linear, but having your allowance doubled every year is exponential. After 10 years, the difference would be enormous!

⚠

Common Mistake

Students sometimes confuse a steep linear function with an exponential one. A line like y = 100x + 5 rises quickly, but it is still linear because the rate of change is constant (always adding 100). An exponential like y = 5 · 1.5ˣ may start slowly but eventually overtakes any linear function. Always run the numerical tests rather than relying on visual impressions alone.

SECTION 8

Connecting to Advanced Modeling

The skill of choosing between linear and exponential models is actually the first step in a much larger toolkit. In later math courses, you will encounter additional model types and more sophisticated methods for fitting them to data. The table below gives you a preview of how the concepts you are learning now connect to what comes next.

What You Learn Now	What Comes Later
Visual inspection of scatter plots	Residual plots to assess model fit quantitatively
First differences test for linearity	Correlation coefficient (r) and R² values
Common ratio test for exponential	Logarithmic transformations to linearize exponential data
Linear or exponential or neither	Quadratic, logarithmic, sinusoidal, logistic, and polynomial models
Justifying with numerical evidence	Formal hypothesis testing and regression analysis

Even though these advanced tools are more powerful, they all build on the same fundamental question you are answering now: "What pattern does the data follow?" Mastering the linear-vs-exponential decision gives you a strong foundation for every modeling challenge you will face in Algebra 2, Pre-Calculus, Statistics, and beyond.

SECTION 9

Practice Problems

Question 1: CONCEPTUAL

Explain in your own words how the first differences test and the common ratio test help you distinguish between a linear and an exponential model. Why is it important to check that the x-values are equally spaced before using these tests?

Show answer

The first differences test calculates the change between consecutive y-values (yₙ₊₁ − yₙ). If these differences are roughly constant, the output is growing by addition, which is the hallmark of a linear model y = mx + b. The common ratio test divides consecutive y-values (yₙ₊₁ ÷ yₙ). If these ratios are roughly constant, the output is growing by multiplication, which is the hallmark of an exponential model y = a · bˣ. Both tests assume equal steps in x because the "constant" difference or ratio corresponds to a single unit step. If x-steps are unequal, the differences and ratios will vary even for perfectly linear or exponential data, potentially leading to a wrong classification.

Question 2: BASIC CALCULATION

A data set has x-values 1, 2, 3, 4, 5 and corresponding y-values 7, 12, 17, 22, 27. Compute the first differences and common ratios. Determine whether the data is best modeled by a linear, exponential, or neither model.

Show answer

First differences: 12 − 7 = 5, 17 − 12 = 5, 22 − 17 = 5, 27 − 22 = 5. All first differences equal 5 — they are constant. Common ratios: 12 ÷ 7 ≈ 1.71, 17 ÷ 12 ≈ 1.42, 22 ÷ 17 ≈ 1.29, 27 ÷ 22 ≈ 1.23. The ratios are not constant. Since the first differences are constant at 5, this data is best modeled by a linear equation. With slope m = 5 and y-intercept b = 2 (since y = 5(1) + 2 = 7), the model is y = 5x + 2.

Question 3: INTERMEDIATE

Consider the data set: x = 0, 1, 2, 3, 4 and y = 500, 400, 320, 256, 204.8. (a) Compute first differences and common ratios. (b) Classify the model. (c) Write the equation and predict y when x = 6.

Show answer

(a) First differences: 400 − 500 = −100, 320 − 400 = −80, 256 − 320 = −64, 204.8 − 256 = −51.2. These are not constant. Common ratios: 400 ÷ 500 = 0.8, 320 ÷ 400 = 0.8, 256 ÷ 320 = 0.8, 204.8 ÷ 256 = 0.8. All ratios equal 0.8 — constant. (b) The data follows an exponential model (specifically exponential decay, since the ratio is between 0 and 1). (c) The equation is y = 500 · 0.8ˣ. For x = 6: y = 500 · 0.8⁶ = 500 · 0.262144 = 131.072. The predicted y-value when x = 6 is approximately 131.07.

Question 4: APPLIED

A local coffee shop tracks its weekly revenue (in dollars) for the first five weeks after opening: Week 1: $1,200; Week 2: $1,450; Week 3: $1,680; Week 4: $1,920; Week 5: $2,150. The owner wants to predict revenue for Week 10. (a) Determine if a linear or exponential model is more appropriate. (b) Write the equation and make the prediction.

Show answer

(a) First differences: 1450 − 1200 = 250, 1680 − 1450 = 230, 1920 − 1680 = 240, 2150 − 1920 = 230. These are approximately constant (≈ 237.5 on average). Common ratios: 1450 ÷ 1200 ≈ 1.208, 1680 ÷ 1450 ≈ 1.159, 1920 ÷ 1680 ≈ 1.143, 2150 ÷ 1920 ≈ 1.120. The ratios are not constant (they decrease). Since the first differences are approximately constant at about 237.5, a linear model is more appropriate. (b) Using the average slope m ≈ 237.50 and the first data point (week 1, $1,200): y = 237.5x + 962.5. At week 10: y = 237.5(10) + 962.5 = $3,337.50. The owner can expect revenue of approximately $3,338 in Week 10, assuming the trend continues.

Question 5: CRITICAL THINKING

A student is given the data set x = 0, 1, 2, 3, 4, 5 and y = 10, 15, 22, 34, 45, 61. The student computes first differences (5, 7, 12, 11, 16) and common ratios (1.5, 1.47, 1.55, 1.32, 1.36). She concludes that neither model fits. Her classmate argues that an exponential model is a reasonable approximation. Who is more correct, and why? What additional information or analysis might help settle the disagreement?

Show answer

Both students make valid observations, but the first student is more justified in her conclusion. The first differences vary from 5 to 16, showing no consistency — this rules out a linear model. The common ratios range from 1.32 to 1.55, which is a meaningful spread (about a 17% variation from the average ratio of ≈1.44). While the ratios are closer to being constant than the differences are, the variation is substantial enough to question an exponential fit. The classmate's argument has some merit — if this were real-world data with measurement error, some variation is expected. To settle the disagreement, the students could: (1) plot the data and see if it visually resembles an exponential curve; (2) compute second differences to check for a possible quadratic fit; (3) use a graphing calculator to find R² values for both exponential and quadratic regression models and compare which fits better. In real data analysis, perfect constancy is rare, and judgment is needed to decide if the variation is due to noise or a genuinely different pattern.

SUMMARY

Lesson Summary

Summary

Choosing the right model type is all about recognizing patterns. A linear model (y = mx + b) fits data where the first differences between consecutive y-values are approximately constant — meaning the output changes by the same amount each step. An exponential model (y = a · bˣ) fits data where the common ratios between consecutive y-values are approximately constant — meaning the output is multiplied by the same factor each step. If neither test produces consistent results, the data may follow a different type of model or no simple model is appropriate.

To justify your choice, always show your numerical work — state the computed differences or ratios and explain why they support your classification. Remember to confirm that the x-values are equally spaced before applying these tests. With practice, you will develop an intuition for which model fits just by looking at a table or graph, but the numerical evidence is what makes your reasoning convincing and complete.