Advanced Exponential and Logarithmic Models

Model	General form	Characteristic behaviour
Linear	y = mx + b	Constant rate of change
Quadratic	y = ax² + bx + c	Parabolic relationship
Exponential	y = Abˣ or Aeᵏˣ	Constant multiplicative change
Logarithmic	y = a ln(x) + b	Rapid initial change followed by slowing change
Logistic	y = K/(1 + Ae⁻ʳˣ)	Growth that approaches a limiting value

y = 2x + 1

A straight line with a constant slope.

Linear Models

A linear model describes a relationship with a constant rate of change.

The general equation is:

y = mx + b

where:

• m = slope or rate of change.

• b = y-intercept or initial value.

Characteristics

• Graph is a straight line.

• Equal changes in x produce equal changes in y.

• Slope is constant.

• The model may increase, decrease, or remain constant.

Worked Example 1: Taxi Fare

A taxi company charges a base fare of $5 and $2.50 per kilometre.

The model is:

C(d) = 5 + 2.5d

Calculate the fare for a 14 km journey.

C(14) = 5 + 2.5(14)

C(14) = 5 + 35

C(14) = $40

Interpreting the model

The slope, 2.5, represents the additional cost per kilometre.

The intercept, 5, represents the fixed starting charge.

A linear model is appropriate because the cost increases by a constant amount for every additional kilometre.

When is a linear model appropriate?

Linear models are useful when the relationship has a reasonably constant additive rate of change.

Examples include fixed-rate wages, simple pricing systems, uniform motion, and some short-term trends.

However, linear models may become unrealistic if rates change, limits are reached, or the underlying relationship is nonlinear.

Quadratic Models

A quadratic model describes a relationship involving a squared variable.

The general equation is:

y = ax² + bx + c, where a ≠ 0.

Its graph is a parabola.

Characteristics

• Graph is U-shaped or inverted U-shaped.

• May have a maximum or minimum.

• Rate of change is not constant.

• Second differences are constant for equally spaced x-values.

• Often describes relationships involving acceleration, area, or optimisation.

Worked Example 2: Projectile Motion

A ball is thrown upward from a height of 2 metres.

Its height is modelled by:

h(t) = −5t² + 20t + 2

where t is measured in seconds.

Find the maximum height.

For a quadratic function, the vertex occurs at:

t = −b/(2a)

Here:

a = −5

b = 20

Therefore:

t = −20/[2(−5)]

t = 2 seconds

Substitute:

h(2) = −5(2²) + 20(2) + 2

h(2) = −20 + 40 + 2

Maximum height = 22 m

The quadratic model is appropriate because the ball experiences approximately constant downward acceleration.

Why not use a linear model?

A linear model assumes constant vertical velocity.

The ball slows while rising, stops momentarily at its maximum height, and accelerates downward.

Its vertical velocity changes continuously, so a quadratic height–time model is more appropriate when air resistance is neglected.

Exponential Models

Exponential models describe processes with constant proportional change.

Common forms are:

y = Abˣ

and

y = Aeᵏˣ

where A is the initial quantity.

Characteristics

• Equal increases in x produce equal multiplication factors.

• The absolute rate of change depends on the current quantity.

• Growth is unbounded when the growth rate remains positive.

• Decay approaches zero when the decay constant is negative.

• Often describes compound interest, early population growth, and first-order decay.

Worked Example 3: Bacterial Growth

A bacterial population begins with 300 organisms and doubles every 4 hours.

The model is:

P(t) = 300 × 2^(t/4)

Calculate the population after 12 hours.

P(12) = 300 × 2^(12/4)

P(12) = 300 × 2³

P(12) = 300 × 8

P(12) = 2,400 bacteria

Why is an exponential model appropriate?

The population multiplies by the same factor over each four-hour interval.

A linear model would assume the same number of bacteria is added every four hours, rather than the population doubling.

Important limitation

Exponential growth is often suitable only while resources remain sufficiently abundant.

Eventually, competition, nutrient depletion, or other environmental constraints may reduce the growth rate.

Logarithmic Models

A logarithmic model describes a relationship that changes rapidly at first but then changes more slowly.

A common form is:

y = a ln(x) + b, where x > 0.

The function ln(x) is the natural logarithm.

Characteristics

• Domain is restricted to positive x-values in the basic form.

• When a > 0, the graph increases at a decreasing rate.

• When a < 0, the graph decreases at a decreasing absolute rate.

• The basic graph has a vertical asymptote at x = 0.

• Logarithmic functions are inverses of exponential functions.

Worked Example 4: Learning and Experience

A simplified model for a worker's performance score is:

S(x) = 40 + 12ln(x)

where x represents the number of completed practice sessions, with x ≥ 1.

Calculate the predicted score after 10 sessions.

S(10) = 40 + 12ln(10)

Since ln(10) ≈ 2.3026:

S(10) ≈ 40 + 12(2.3026)

S(10) ≈ 40 + 27.6312

S(10) ≈ 67.63

Interpreting the model

The score increases as the number of sessions increases, but the improvement per additional session becomes smaller.

This illustrates diminishing returns.

However, the logarithmic model does not have a finite upper limit. It would eventually predict arbitrarily large scores if extended indefinitely.

If performance has a strict maximum, another model may be more suitable for long-term predictions.

Logistic Models

A logistic model describes growth that slows as a quantity approaches a limiting value.

A common form is:

P(t) = K/(1 + Ae⁻ʳᵗ)

where:

• K = carrying capacity or limiting value.

• A = constant determined by the initial quantity.

• r = intrinsic growth parameter.

• t = time.

Characteristics

• Usually produces an S-shaped graph when starting well below K.

• Initially resembles exponential growth.

• Growth slows as the population approaches carrying capacity.

• Maximum instantaneous growth occurs at P = K/2 in the standard model.

• The population approaches K as time increases.

Worked Example 5: Wildlife Population

A wildlife population follows:

P(t) = 2000/(1 + 19e⁻⁰·⁴ᵗ)

Calculate the population after 10 years.

P(10) = 2000/(1 + 19e⁻⁴)

Since e⁻⁴ ≈ 0.0183156:

P(10) ≈ 2000/[1 + 19(0.0183156)]

P(10) ≈ 2000/1.347997

P(10) ≈ 1,484 organisms

Interpreting the model

The carrying capacity is 2,000 organisms.

The initial population is:

P(0) = 2000/(1 + 19)

P(0) = 100

The model predicts that the population increases from 100 and gradually approaches 2,000.

A logistic model is appropriate when population growth is limited by environmental resources.

Identifying the Most Appropriate Model

One of the most important modelling skills is recognising which mathematical relationship best describes a situation.

Recognising patterns in data

These patterns are useful clues, but they are not sufficient by themselves to establish which model is correct.

The underlying context and assumptions must also be considered.

Worked Example 6: Identifying a Model from a Table

Consider the following data.

Calculate the first differences:

9 − 5 = 4

13 − 9 = 4

17 − 13 = 4

21 − 17 = 4

The first differences are constant.

Therefore, a linear model is appropriate.

The slope is 4.

Using y = mx + b:

5 = 4(1) + b

b = 1

The model is:

y = 4x + 1

Worked Example 7: Identifying Quadratic Growth

Consider:

First differences:

3, 5, 7, 9

Second differences:

2, 2, 2

The constant second differences suggest a quadratic model.

The equation is:

y = x² + 2x + 2

Check x = 3:

y = 3² + 2(3) + 2

y = 9 + 6 + 2

y = 17

The equation matches the data.

Worked Example 8: Identifying Exponential Growth

Consider:

Calculate successive ratios:

150/100 = 1.5

225/150 = 1.5

337.5/225 = 1.5

506.25/337.5 = 1.5

The ratio is constant.

Therefore, an exponential model is appropriate.

y = 100(1.5)ˣ

The quantity increases by 50% for every one-unit increase in x.

Distinguishing Logarithmic and Logistic Patterns

Logarithmic and logistic models can be difficult to distinguish because both may show growth that gradually slows.

However, their long-term behaviour is fundamentally different.

A logarithmic model such as:

y = 10 + 5ln(x)

continues increasing without a finite upper limit, although its rate of increase becomes smaller.

A logistic model such as:

y = 100/(1 + 9e⁻⁰·⁵ˣ)

approaches a fixed upper limit of 100.

Comparing their behaviour

The distinction matters when making long-term predictions.

For example, a learning-performance score that cannot exceed 100 may be better represented by a bounded model than by a logarithmic function that eventually exceeds 100.

A logistic model is one possible bounded model, although it should only be selected when its shape and assumptions are appropriate.

A Systematic Method for Selecting a Model

Choosing a mathematical model should involve more than looking at a graph and selecting the closest-looking curve.

A reliable process follows several stages.

Step 1: Understand the situation

Ask what the variables represent and what is known about their relationship.

For example, does a price increase by a fixed amount, or does a population increase by a fixed percentage?

Step 2: Examine the data

Construct an appropriate scatter plot and investigate patterns.

Look for straight-line trends, curvature, changing growth rates, or evidence of a limiting value.

Step 3: Identify candidate models

Consider which models are mathematically and scientifically plausible.

Do not assume that the most complicated model is necessarily the best.

Step 4: Fit the models

Estimate the parameters using the available data.

For example, linear regression can estimate a slope and intercept, while nonlinear regression can estimate logistic model parameters.

Step 5: Evaluate model performance

Compare prediction errors and examine residuals.

Step 6: Interpret and validate

Determine whether the selected model makes sense in context and performs adequately on data not used to fit it.

Justifying Model Selection

A strong model justification should contain four elements:

Observed pattern + Mathematical evidence + Contextual reasoning + Limitations

Example: Choosing a Model for Population Growth

Suppose a researcher measures a population over 12 years.

The population grows rapidly at first but begins to level off near 5,000 organisms.

A justified explanation might be:

"A logistic model is appropriate because the population initially increases rapidly, but its growth rate decreases as the population approaches approximately 5,000 organisms. This suggests a carrying-capacity limitation. An exponential model would predict unlimited growth and is therefore less plausible over the long term. However, the logistic model assumes that carrying capacity and the intrinsic growth parameter remain constant."

This explanation is stronger than simply saying:

"The graph looks like an S, so the model is logistic."

The first explanation connects the mathematical pattern to the underlying situation.

Evaluating Model Performance

Different models can be fitted to the same dataset.

The model that produces the smallest error on the training data is not necessarily the most appropriate model.

A good model should also make reliable predictions and have assumptions consistent with the situation.

Residuals

A residual is the difference between an observed value and its predicted value.

Residual = Observed value − Predicted value

In symbols:

eᵢ = yᵢ − ŷᵢ

Where:

• yᵢ = observed value.

• ŷᵢ = predicted value.

• eᵢ = residual.

A positive residual means the model underestimated the observation.

A negative residual means the model overestimated the observation.

Worked Example 9: Calculating Residuals

Suppose a model predicts examination scores.

For Student A:

Residual = 70 − 68 = 2

For Student B:

Residual = 75 − 78 = −3

Residuals help identify patterns the model has failed to explain.

Interpreting residual plots

A suitable model often produces residuals scattered around zero without a systematic pattern.

A curved residual pattern from a linear model may suggest that a quadratic or another nonlinear model would be more appropriate.

However, random-looking residuals alone do not prove that a model is valid.

Measuring Model Accuracy

Several numerical measures help compare model performance.

Mean Absolute Error (MAE)

MAE is the average absolute prediction error.

Mean Squared Error (MSE)

MSE is the average squared prediction error.

Root Mean Squared Error (RMSE)

RMSE is the square root of MSE.

MAE and RMSE are measured in the same units as the response variable.

RMSE penalises larger errors more heavily than MAE.

Coefficient of determination (R²)

R² measures the proportion of variation in the response explained by a model relative to a mean-only baseline, using the usual squared-error definition.

For ordinary least-squares models with an intercept, it typically lies between 0 and 1 on the training data.

However, R² can be negative for some models or when evaluated on new data.

A high R² does not automatically mean a model is appropriate.

A model can fit observed data well while making poor predictions outside the observed range.

Worked Example 10: Comparing Two Models

Suppose a researcher fits a linear model and a quadratic model to the same dataset.

The models produce the following residuals on an independent validation dataset.

Calculate RMSE for each model.

Linear model

Square the residuals:

4² = 16

(−3)² = 9

5² = 25

(−4)² = 16

2² = 4

Sum:

16 + 9 + 25 + 16 + 4 = 70

MSE = 70/5 = 14

RMSE = √14

RMSE ≈ 3.74

Quadratic model

Square the residuals:

2² = 4

(−2)² = 4

1² = 1

(−2)² = 4

1² = 1

Sum:

4 + 4 + 1 + 4 + 1 = 14

MSE = 14/5 = 2.8

RMSE = √2.8

RMSE ≈ 1.67

Interpretation

The quadratic model has a lower validation RMSE.

It therefore makes more accurate predictions on this validation dataset according to RMSE.

This supports selecting the quadratic model, provided its assumptions and contextual interpretation are also reasonable.

We should still investigate whether the improvement is consistent across other data and whether the model generalises beyond the observed range.

Overfitting and Underfitting

Two important problems can occur when selecting models.

Underfitting

Underfitting occurs when a model is too simple to capture important patterns.

For example, a straight line may fit poorly when the underlying relationship is strongly curved.

Overfitting

Overfitting occurs when a model captures random variation or noise in the training data rather than the underlying relationship.

For example, a high-degree polynomial might pass through every observed point but behave unpredictably between or beyond those points.

Balancing complexity and accuracy

The goal is not to find the most complicated model.

It is to find a model that captures the important relationship while making reliable predictions.

This is why model validation matters.

A common approach is to fit models using training data and compare their performance using separate validation data.

Interpolation and Extrapolation

Mathematical models are often used to estimate values that were not directly observed.

Interpolation

Interpolation means predicting within the range of the observed data.

For example, if a model was developed using temperatures between 10°C and 30°C, estimating a value at 20°C is interpolation.

Extrapolation

Extrapolation means predicting outside the observed range.

Estimating a value at 60°C using the same model would be extrapolation.

Extrapolation is generally more uncertain because the relationship may change outside the observed conditions.

Example: Population Predictions

Suppose a population has been observed for five years and has increased from 100 to 300.

An exponential model may fit these early observations well.

However, predicting the population 100 years into the future could be unrealistic if food and space are limited.

A logistic model might be more plausible, but estimating its carrying capacity from only the early growth phase can also be highly uncertain.

A model's reliability depends on both its mathematical fit and the conditions under which it is used.

Real-World Applications of Model Selection

Authentic Modelling Investigation: Selecting the Best Population Model

A wildlife researcher records the population of an introduced species over several years.

The following hypothetical data are collected.

The researcher wants to select a model for future population predictions.

Step 1: Examine the pattern

The population increases throughout the observation period.

However, the increases over successive two-year intervals are:

82, 121, 148, 148, 122, 85

The increases initially become larger and then smaller.

This suggests that the population's absolute growth rate is beginning to slow.

Step 2: Consider possible models

A linear model would assume a constant increase per year.

An exponential model would assume a constant proportional growth rate.

A logistic model allows growth to accelerate initially and then slow as carrying capacity is approached.

The logistic model is a plausible candidate.

Step 3: Propose a logistic model

Suppose the researcher estimates a carrying capacity of 900 organisms.

A candidate model is:

P(t) = 900/(1 + 8e⁻⁰·³ᵗ)

The initial population is:

P(0) = 900/(1 + 8)

P(0) = 100

The model reproduces the initial observation.

Step 4: Compare predictions with observations

The model captures the general S-shaped pattern but systematically underestimates the later observations.

This suggests that the proposed parameters require adjustment.

It would be inappropriate to claim that the logistic model is accurate simply because its graph has the expected shape.

Step 5: Improve the investigation

The researcher could:

• Fit the logistic parameters using nonlinear regression rather than choosing them manually.

• Compare the fitted logistic model with exponential and other plausible models.

• Examine residuals for systematic patterns.

• Evaluate predictions using later observations.

• Investigate whether the estimated carrying capacity is biologically realistic.

Step 6: Communicate a justified conclusion

"The observed population initially experienced increasing absolute growth, followed by smaller increases in later years. This pattern is consistent with a logistic-type growth process. However, the proposed logistic equation systematically underestimated the observations, so its parameters should be re-estimated and its predictions validated before it is used for management decisions."

This is an example of critical mathematical modelling: the researcher considers both the observed pattern and the quality of the fitted equation.

Common Mistakes

Mistake 1: Selecting a model based only on its appearance

Different models can produce similar curves over short intervals.

Correction: Examine mathematical patterns, assumptions, and residuals.

Mistake 2: Assuming the model with the highest R² is always best

A complicated model may fit training data well but predict poorly.

Correction: Consider validation performance, model complexity, and contextual plausibility.

Mistake 3: Confusing linear and exponential growth

Linear growth involves constant additions. Exponential growth involves constant multiplication factors.

Correction: Compare differences and ratios.

Mistake 4: Confusing logarithmic and logistic growth

A logarithmic function continues increasing without a finite upper bound, while a standard increasing logistic function approaches carrying capacity.

Correction: Examine long-term behaviour.

Mistake 5: Ignoring the domain

Some models are mathematically defined for values that make no sense in the situation.

Correction: State meaningful restrictions on the independent variable.

Mistake 6: Treating model predictions as exact

Predictions are estimates that depend on assumptions and data quality.

Correction: Communicate uncertainty and limitations.

Mistake 7: Extrapolating too far

A model may work within the observed range but fail outside it.

Correction: Be cautious when making long-term predictions.

Mistake 8: Ignoring the meaning of parameters

A mathematical equation is more useful when its coefficients can be interpreted in context.

Correction: Explain slopes, initial values, growth rates, turning points, and carrying capacities.

Key Terms

• Mathematical Model: A mathematical representation of a real-world relationship or system.

• Linear Model: A model with a constant additive rate of change.

• Quadratic Model: A model containing a squared variable, producing a parabolic graph.

• Exponential Model: A model with a constant proportional rate of change.

• Logarithmic Model: A model involving a logarithmic relationship.

• Logistic Model: A model describing growth that approaches a limiting value.

• Carrying Capacity: The limiting population size in a standard logistic growth model.

• Slope: The rate of change of a linear function.

• Vertex: The maximum or minimum point of a parabola.

• Growth Rate: A measure of how quickly a quantity increases.

• Residual: The difference between an observed value and a predicted value.

• Mean Absolute Error: The average absolute prediction error.

• Root Mean Squared Error: The square root of the average squared prediction error.

• Coefficient of Determination (R²): A measure comparing a model's squared prediction errors with those of a mean-only baseline.

• Underfitting: Failure to capture important patterns because a model is too simple.

• Overfitting: Fitting random variation rather than the underlying relationship.

• Interpolation: Prediction within the observed data range.

• Extrapolation: Prediction outside the observed data range.

• Model Validation: Evaluating how well a model performs on data not used to fit it.

• Model Assumption: A condition under which a mathematical model is intended to apply.

• Model Limitation: A restriction on the accuracy, usefulness, or applicability of a model.

Key Takeaways

• Linear models describe constant additive change.

• Quadratic models describe parabolic relationships and can contain maxima or minima.

• Exponential models describe constant proportional change.

• Logarithmic models can describe rapidly changing relationships that gradually slow.

• Logistic models describe growth that approaches a carrying capacity.

• Differences, second differences, ratios, and graphical patterns help identify candidate models.

• Model selection should consider mathematical fit and real-world context.

• Residuals and error measures help evaluate model performance.

• A more complicated model is not automatically a better model.

• Predictions outside the observed range require particular caution.

• Every mathematical model depends on assumptions.

• A justified model selection explains why the model is appropriate and acknowledges its limitations.

Check Your Understanding

Questions

Question 1: Identifying Models

Identify the most appropriate model type for each situation.

a) A company charges $15 plus $4 per item delivered.

b) A ball is thrown vertically upward.

c) A population doubles every three years.

d) A quantity increases rapidly at first, then more slowly, without an assumed finite limit.

e) A population grows toward an environmental carrying capacity.

Question 2: Linear Modelling

A company charges a fixed fee of $12 plus $3 per kilometre.

Write a model and calculate the cost for 15 km.

Question 3: Quadratic Modelling

A projectile's height is:

h(t) = −5t² + 30t + 2

Find the time at which the projectile reaches its maximum height and calculate that height.

Question 4: Exponential Modelling

A population begins with 200 organisms and increases by 25% each year.

Write a model and calculate the population after four years.

Question 5: Logarithmic Modelling

A simplified performance model is:

S(x) = 30 + 10ln(x)

Calculate S(5), correct to two decimal places.

Question 6: Logistic Modelling

A population follows:

P(t) = 3000/(1 + 14e⁻⁰·²ᵗ)

Determine the carrying capacity and initial population.

Question 7: First Differences

Determine the most appropriate model for:

Question 8: Second Differences

Determine the most appropriate model for:

Question 9: Exponential Ratios

Determine the most appropriate model for:

Write its equation.

Question 10: Model Accuracy

Two models have validation RMSE values of 4.2 and 2.8.

Which model performs better according to RMSE, and what additional factors should be considered?

Question 11: Residuals

A model predicts 75, but the observed value is 82.

Calculate the residual and interpret it.

Question 12: Extrapolation

A model is developed using data collected over five years.

Explain why predicting 50 years into the future may be unreliable.

Question 13: Model Selection

A population increases rapidly at first, but observations suggest that it is approaching approximately 10,000 organisms.

Explain why a logistic model may be more appropriate than an exponential model.

Question 14: Critical Interpretation

A researcher fits a high-degree polynomial that passes through every observed data point.

Explain why this does not guarantee that the model will make accurate future predictions.

Question 15: Integrated Application

A business records the following sales:

a) Identify a suitable model.

b) Determine the equation.

c) Predict sales in month 5.

d) Explain one limitation of the model.

Fully Worked Answers

Answer 1: Identifying Models

a) Linear — constant additional charge per item.

b) Quadratic — vertical height under approximately constant gravitational acceleration.

c) Exponential — constant doubling factor.

d) Logarithmic — rapid initial increase followed by slower growth without a fixed upper bound.

e) Logistic — growth limited by carrying capacity.

Answer 2: Linear Modelling

The fixed fee is $12, and the charge per kilometre is $3.

C(d) = 12 + 3d

For d = 15:

C(15) = 12 + 3(15)

C(15) = 12 + 45

C(15) = $57

Answer 3: Quadratic Modelling

Given:

h(t) = −5t² + 30t + 2

The maximum occurs at:

t = −b/(2a)

t = −30/[2(−5)]

t = 3 seconds

Substitute:

h(3) = −5(3²) + 30(3) + 2

h(3) = −45 + 90 + 2

Maximum height = 47 m

Answer 4: Exponential Modelling

Initial population = 200.

Growth factor = 1.25.

Model:

P(t) = 200(1.25)ᵗ

After four years:

P(4) = 200(1.25)⁴

P(4) = 200(2.44140625)

P(4) ≈ 488 organisms

Answer 5: Logarithmic Modelling

S(x) = 30 + 10ln(x)

Substitute x = 5:

S(5) = 30 + 10ln(5)

S(5) ≈ 30 + 10(1.609438)

S(5) ≈ 46.09438

S(5) ≈ 46.09

Answer 6: Logistic Modelling

Given:

P(t) = 3000/(1 + 14e⁻⁰·²ᵗ)

The carrying capacity is the numerator:

K = 3000

At t = 0:

P(0) = 3000/(1 + 14)

P(0) = 3000/15

P₀ = 200

Answer 7: First Differences

First differences:

12 − 7 = 5

17 − 12 = 5

22 − 17 = 5

The first differences are constant, so the model is linear.

Slope = 5.

Using y = 5x + b and the point (1, 7):

7 = 5 + b

b = 2

y = 5x + 2

Answer 8: Second Differences

First differences:

3, 5, 7, 9

Second differences:

2, 2, 2

The second differences are constant, so a quadratic model is appropriate.

The equation is:

y = x² + 2x + 3

Check x = 3:

y = 9 + 6 + 3 = 18

The model matches the table.

Answer 9: Exponential Ratios

Successive ratios are:

100/50 = 2

200/100 = 2

400/200 = 2

The constant ratio indicates exponential growth.

Initial value = 50.

Growth factor = 2.

y = 50(2)ˣ

Answer 10: Model Accuracy

The model with RMSE = 2.8 performs better according to validation RMSE because it has a smaller typical squared-error-based prediction magnitude.

However, we should also consider model complexity, residual patterns, assumptions, interpretability, and performance on other representative data.

Answer 11: Residuals

Residual = Observed − Predicted

Residual = 82 − 75

Residual = +7

The positive residual indicates that the model underestimated the observation by 7 units.

Answer 12: Extrapolation

Predicting 50 years into the future using only five years of observations involves substantial extrapolation.

The relationship may change because of environmental, economic, technological, or other factors.

Therefore, the model may no longer represent the underlying process accurately.

Answer 13: Model Selection

A logistic model is appropriate because it allows population growth to slow as the population approaches carrying capacity.

An exponential model would predict unlimited growth if its positive relative growth rate remained constant.

A logistic model can represent the apparent limiting population of approximately 10,000.

However, additional observations and ecological evidence would help confirm whether this carrying capacity is realistic.

Answer 14: Critical Interpretation

A high-degree polynomial may fit every observed data point but also capture random variation or measurement noise.

This is called overfitting.

The polynomial may behave unpredictably between observations or outside the observed range.

Accurate training-data fit does not guarantee accurate future predictions.

The model should be evaluated using separate validation data and by examining whether its behaviour makes sense.

Answer 15: Integrated Application

Part a: Identify the model.

Calculate ratios:

180/120 = 1.5

270/180 = 1.5

405/270 = 1.5

The ratio is constant, so an exponential model is appropriate.

Part b: Determine the equation.

At month 1, sales are 120.

The model can be written as:

S(m) = 120(1.5)^(m − 1)

Part c: Predict month 5.

S(5) = 120(1.5)⁴

S(5) = 120(5.0625)

S(5) = 607.5

The model predicts approximately 608 sales if sales are counted in whole units.

Part d: Explain a limitation.

The model assumes that sales continue increasing by 50% every month.

This may be unrealistic because demand, market size, competition, and available resources can limit growth.

The model may therefore be appropriate for short-term predictions but unreliable over longer periods.

Final Reflection

Comparing mathematical models is an essential part of mathematical reasoning because different relationships require different mathematical descriptions.

Linear, quadratic, exponential, logarithmic, and logistic models each capture different patterns of change.

Understanding their equations and graphs allows us to recognise these patterns, but selecting an appropriate model also requires evaluating the context, assumptions, accuracy, and limitations.

The best mathematical model is not necessarily the most complicated or the one that fits the observed data most closely. It is the model that provides a useful, defensible representation of the situation and produces reliable predictions for its intended purpose.

This ability to compare, justify, interpret, and critically evaluate models is fundamental to mathematics, statistics, science, engineering, economics, and evidence-based decision-making.