The Central Idea
A mathematical model is a representation of relationships and mechanisms. The form of that representation should match what the system is doing.
A single relationship might be expressed as
\[ d=vt. \]An input-output rule might be expressed as
\[ C(q)=500+8q. \]Several interacting relationships might require a system:
\[ \begin{cases} x+y=100,\\ 3x+5y=420. \end{cases} \]A process that evolves through repeated steps may be better described by an algorithm.
The modeling question is therefore not merely “Which formula do I know?” but “Which mathematical structure matches the mechanism?”
Four Common Modeling Structures
In introductory modeling, four broad structures appear repeatedly.
- Equations express conditions that quantities must satisfy.
- Functions describe how outputs depend on inputs.
- Systems represent several relationships that must hold together.
- Algorithms specify a sequence of computational or decision steps.
These categories overlap. A function may be defined by an equation, a system may contain functions, and an algorithm may repeatedly evaluate equations or functions.
Start with the Mechanism
Before selecting notation, ask what actually connects the quantities.
Common mechanisms include:
- constant addition;
- constant multiplication;
- proportionality;
- conservation or balance;
- capacity limitations;
- feedback;
- competition;
- accumulation;
- random variation;
- repeated decision rules.
Once the mechanism is clear, the mathematical structure becomes easier to choose.
When a Single Equation Is Enough
A single equation is often appropriate when one condition directly relates the relevant quantities.
For example, if distance is speed multiplied by time,
\[ d=vt. \]If total cost equals fixed cost plus variable cost,
\[ C=F+cq. \]If a rectangle has fixed perimeter \(P\),
\[ 2L+2W=P. \]Each equation expresses one central relationship.
Equations as Conditions
An equation does not always define one variable explicitly as a function of another.
Consider
\[ x^2+y^2=25. \]This equation describes all points on a circle of radius \(5\).
It is a model of a geometric condition:
\[ \text{distance from the origin}=5. \]The equation is useful even though the entire circle is not one function \(y=f(x)\).
Equations from Conservation
Many models begin with a balance principle.
A general conservation structure is
\[ \text{new amount} = \text{old amount} + \text{inflow} - \text{outflow}. \]For inventory,
\[ I_{t+1}=I_t+Q_t-D_t. \]For a bank account,
\[ B_{t+1}=B_t+D_t-W_t. \]Conservation laws often provide the first equation in a model.
When a Function Is the Natural Choice
A function is especially useful when the main question is how one quantity depends on another.
For example,
\[ C(q)=1000+6q \]makes the dependency explicit:
- input: production quantity \(q\);
- output: total cost \(C\).
Function notation emphasizes the input-output structure of the model.
Function Notation Clarifies Dependency
Compare
\[ C=1000+6q \]with
\[ C(q)=1000+6q. \]The two expressions may encode the same numerical relationship, but the second makes the dependency more visible.
It tells us that \(C\) is being treated as a function of \(q\).
Choosing the Independent Variable
A model may contain several quantities, but one may be selected as the main input.
If position changes over time, we might write
\[ s(t). \]If demand changes with price, we might write
\[ D(p). \]If cost changes with production,
\[ C(q). \]The notation should reflect the modeling question.
A Function Can Have Several Inputs
Many outputs depend on more than one quantity.
For example,
\[ Q=f(K,L) \]might model output \(Q\) as a function of capital \(K\) and labor \(L\).
Temperature might be modeled as
\[ T=T(x,y,t), \]depending on position and time.
A function does not have to have only one input.
Linear Functions
Use a linear or affine model when equal changes in the input produce approximately equal changes in the output.
\[ y=mx+b. \]Here \(m\) is the constant rate of change.
Examples include:
- hourly wages with a fixed hourly rate;
- distance traveled at constant speed;
- simple depreciation over a limited period;
- cost with a fixed base plus constant unit cost.
Proportional Models
If one quantity is directly proportional to another, use
\[ y=kx. \]The constant \(k\) is the proportionality constant.
This is a special linear model with zero intercept.
The distinction matters. Not every linear relationship is proportional.
Power Models
Scaling relationships often take the form
\[ y=ax^b. \]For example, the area of a circle depends on the square of radius:
\[ A(r)=\pi r^2. \]The volume of a geometrically similar object often scales with the cube of a characteristic length.
Power models are useful when percentage or scale changes follow a consistent exponent pattern.
Exponential Models
When a quantity changes by a constant multiplicative factor over equal intervals, an exponential model may be appropriate.
\[ P(t)=P_0a^t \]or
\[ P(t)=P_0e^{rt}. \]Examples include idealized compound growth, radioactive decay, and some early-stage population models.
The mechanism is repeated multiplication rather than repeated addition.
Logistic Models
Exponential growth may be unrealistic when resources are limited.
A logistic model introduces a carrying capacity \(K\):
\[ \frac{dP}{dt} = rP\left(1-\frac{P}{K}\right). \]The growth mechanism now depends on both the current population and the remaining capacity.
The model structure should change when the mechanism changes.
Piecewise Functions
Some systems follow different rules in different regions.
Suppose shipping costs \(\$8\) for orders below \(\$50\) and is free otherwise:
\[ S(x)= \begin{cases} 8, & x<50,\\ 0, & x\geq 50. \end{cases} \]A piecewise function is more faithful than forcing one formula to describe both regimes.
When One Equation Is Not Enough
Many real systems contain several quantities that interact.
Suppose a company produces products \(x\) and \(y\), and both labor and material are limited.
The model may require
\[ 2x+y\leq 100 \]and
\[ x+3y\leq 120. \]No single equation captures both restrictions. A system is required.
Systems of Equations
A system of equations represents several conditions that must be satisfied simultaneously.
For example,
\[ \begin{cases} x+y=50,\\ 12x+20y=760. \end{cases} \]One equation might represent total quantity, while the other represents total cost.
The solution must satisfy both relationships at the same time.
Systems of Inequalities
Optimization and resource-allocation problems often use systems of inequalities.
\[ \begin{cases} x\geq 0,\\ y\geq 0,\\ 3x+2y\leq 180,\\ x+4y\leq 160. \end{cases} \]These inequalities define the feasible set.
A later objective function can then be optimized over that set.
Coupled Dynamic Systems
When several quantities change over time and influence each other, a dynamic system may be needed.
A simple two-variable model might have
\[ \frac{dx}{dt}=f(x,y) \]and
\[ \frac{dy}{dt}=g(x,y). \]The rate of change of each variable depends on the current state of the system.
Such systems appear in ecology, chemistry, economics, engineering, and many other fields.
When an Algorithm Is the Better Model
Some processes are most naturally represented not by one closed-form equation but by a sequence of steps.
An algorithm is a finite, well-defined procedure for transforming inputs into outputs.
Algorithms are especially useful when:
- the process is iterative;
- the next step depends on the current state;
- decisions occur in stages;
- exact formulas are unavailable;
- simulation is required;
- the model contains logical rules as well as arithmetic.
Example: Repeated Growth as an Algorithm
Suppose an account earns \(5\%\) interest per year.
One representation is
\[ A(t)=A_0(1.05)^t. \]Another is the recurrence
\[ A_{t+1}=1.05A_t. \]This recurrence naturally produces an algorithm:
- start with \(A_0\);
- multiply the current balance by \(1.05\);
- store the result as the next balance;
- repeat for the desired number of periods.
The function and the algorithm describe the same idealized mechanism in different forms.
Recurrence Relations
Recurrence relations are often a bridge between equations and algorithms.
Consider
\[ P_{t+1}=P_t+rP_t. \]This says that the next population equals the current population plus growth.
To generate values, repeatedly apply the same update rule.
State-Update Models
A broad algorithmic modeling pattern is
\[ \text{new state} = \text{update rule}(\text{current state},\text{inputs}). \]In symbols,
\[ x_{t+1}=F(x_t,u_t). \]Here \(x_t\) is the current state and \(u_t\) is an input or decision.
This structure appears in control systems, inventory, finance, population models, games, simulations, and computing.
Algorithms Can Include Conditional Logic
Real processes often include rules of the form:
If this condition is true, do one thing; otherwise, do something else.
For example, an inventory policy might be:
- observe current inventory \(I_t\);
- if \(I_t<200\), order \(500-I_t\) units;
- otherwise, order \(0\);
- update inventory after demand occurs.
This is more naturally expressed as an algorithm than as one elementary formula.
Closed Form versus Iterative Form
Some models have both a direct formula and an iterative representation.
For arithmetic growth,
\[ x_n=x_0+nd \]is a closed form.
The corresponding recurrence is
\[ x_{n+1}=x_n+d. \]Closed forms are convenient for direct calculation. Iterative forms are often closer to the underlying process.
Continuous versus Discrete Models
The structure should also match how the system changes.
A continuously changing quantity may be modeled with a differential equation:
\[ \frac{dP}{dt}=rP. \]A quantity updated once per period may be modeled discretely:
\[ P_{t+1}=(1+r)P_t. \]These models are related but not identical. The time structure is part of the model.
Deterministic versus Stochastic Structure
If the same inputs always produce the same outputs, the model is deterministic.
For example,
\[ C(q)=100+5q \]is deterministic.
If random variation is essential, the model may include a random variable:
\[ Y=\beta_0+\beta_1x+\varepsilon. \]Here \(\varepsilon\) represents unexplained or random variation.
The next lesson develops this distinction more systematically.
Static versus Dynamic Structure
A static model describes relationships at a point or under a fixed set of conditions.
For example,
\[ Q=100-2P \]may describe demand at a given moment.
A dynamic model describes how the system changes over time:
\[ Q_{t+1}=Q_t+\Delta Q_t. \]The following lesson will examine static and dynamic models directly.
Use the Simplest Adequate Structure
More complicated mathematics does not automatically produce a better model.
If a constant-rate relationship is adequate, then
\[ y=mx+b \]may be preferable to a high-degree polynomial.
If two constraints determine the problem, a system of two equations may be preferable to a large simulation.
Complexity should be earned by the mechanism or by the required accuracy.
Mechanistic Models
A mechanistic model is built from an explanation of how the system works.
For example,
\[ \text{change in inventory} = \text{deliveries} - \text{sales} \]leads naturally to
\[ I_{t+1}=I_t+Q_t-D_t. \]The equation comes from a balance mechanism.
Empirical Models
An empirical model is selected mainly because it describes observed data effectively.
For example, measurements may suggest
\[ y\approx a+bx+cx^2. \]This polynomial may fit the data even when it is not derived from a known physical mechanism.
Mechanistic and empirical modeling can also be combined.
Dimensional Reasoning Helps Select Equations
Units can rule out inappropriate model forms.
Suppose distance \(d\) is measured in kilometers, speed \(v\) in kilometers per hour, and time \(t\) in hours.
Then
\[ vt \]has units
\[ \frac{\text{km}}{\text{hour}}\cdot \text{hour} = \text{km}, \]so
\[ d=vt \]is dimensionally consistent.
Units Can Reject a Bad Model
Suppose someone proposes
\[ d=v+t. \]This adds kilometers per hour to hours.
The units do not match, so the proposed equation cannot represent distance without additional structure or conversion.
Dimensional analysis is therefore a fast screening tool during model selection.
Choose Structure Before Solving
A common mistake is to start manipulating symbols before deciding what the model should represent.
A better sequence is:
- identify the quantities;
- identify the mechanism;
- state assumptions and constraints;
- choose the mathematical structure;
- then solve, simulate, estimate, or optimize.
Model selection comes before calculation.
Example: Transportation Cost
A delivery service charges a fixed dispatch fee of \(\$30\) plus \(\$1.80\) per mile.
The mechanism is fixed cost plus constant variable cost, so a linear function is natural:
\[ C(d)=30+1.80d. \]A system or algorithm would be unnecessary unless additional rules are introduced.
Example: Mixture Problem
Suppose \(x\) liters of a \(20\%\) solution and \(y\) liters of a \(50\%\) solution are mixed to make \(100\) liters of a \(32\%\) solution.
Total volume gives
\[ x+y=100. \]Solute balance gives
\[ 0.20x+0.50y=32. \]Two independent conditions are involved, so a system is the natural model.
Example: Tiered Pricing
Suppose electricity costs \(0.12\) dollars per kilowatt-hour for the first \(500\) kWh and \(0.18\) dollars per kilowatt-hour beyond that.
A piecewise function is appropriate:
\[ C(x)= \begin{cases} 0.12x, & 0\leq x\leq 500,\\ 60+0.18(x-500), & x>500. \end{cases} \]The pricing mechanism changes at a threshold, so one unrestricted linear formula would be misleading.
Example: Loan Balance
Suppose a balance earns periodic interest and then receives a fixed payment.
A recurrence is natural:
\[ B_{t+1}=(1+r)B_t-P. \]The process has a repeating order:
- apply interest;
- subtract the payment;
- use the result as the next balance.
This iterative mechanism naturally supports algorithmic computation.
Example: Production Planning
Suppose a firm chooses quantities \(x\) and \(y\) of two products.
Profit is
\[ P=40x+55y. \]Resource constraints are
\[ 2x+3y\leq 180 \]and
\[ 4x+y\leq 160. \]The model therefore needs an objective function plus a system of constraints.
Example: Queue Simulation
A waiting-line system may depend on random arrival times, service times, server availability, and queue order.
A single closed-form equation may not capture all operational details.
An algorithmic simulation can instead:
- generate or observe the next arrival;
- check server availability;
- place customers in the queue if necessary;
- process service completions;
- record waiting times;
- repeat.
The algorithm mirrors the sequence of events in the system.
Hybrid Models
Many serious models combine several structures.
A transportation model might use:
- equations for travel time;
- functions for cost;
- systems of constraints for capacity;
- algorithms for route selection.
The goal is not to force the entire model into one category.
Use each structure where it best represents the relevant part of the mechanism.
Common Error: Choosing a Formula by Appearance
Seeing two quantities does not automatically mean the relationship is linear.
Seeing repeated time periods does not automatically mean exponential growth.
Seeing several variables does not automatically mean a system is required.
Model form should be justified by the mechanism, assumptions, data, or known structure.
Common Error: Using More Variables Than Necessary
Every additional variable increases the information the model must track.
If one variable can be eliminated using
\[ x+y=100, \]then writing
\[ y=100-x \]may reduce a two-variable problem to one variable.
Simplification can improve transparency when it does not remove important structure.
Common Error: Forcing a Closed-Form Formula
Some systems are naturally iterative or computational.
Trying to force every process into a compact symbolic formula can make the model less clear or less faithful to the mechanism.
Recurrences, numerical methods, and simulations are legitimate mathematical models.
Common Error: Ignoring Constraints During Structure Selection
A function may describe an objective but not the full decision problem.
For example,
\[ P(x,y)=40x+50y \]does not by itself represent a production-planning problem if resources are limited.
The constraints are part of the model and may require a system of inequalities.
Common Error: Treating Fit as Mechanism
A curve can fit data well without representing the true process that generated the data.
A high-degree polynomial may pass through many observed points while producing implausible behavior outside the observed range.
Good numerical fit and good mechanistic explanation are different standards.
A Practical Selection Guide
Use a single equation when one central condition relates the quantities.
Use a function when dependency from inputs to outputs is central.
Use a system when several relationships or constraints must hold simultaneously.
Use an algorithm when the process unfolds through repeated, conditional, sequential, or computational steps.
Use a hybrid model when the system genuinely requires more than one of these structures.
A Reliable Procedure
- Identify the quantities and their roles.
- State the modeling question.
- Identify the mechanism connecting the quantities.
- List assumptions, constraints, and domains.
- Ask whether one relationship is sufficient.
- If dependency is central, consider function notation.
- If several conditions must hold together, construct a system.
- If the process is sequential or iterative, consider an algorithm.
- Check dimensional consistency.
- Prefer the simplest structure that adequately represents the mechanism.
- Test whether the chosen structure produces sensible behavior.
- Revise the structure when the mechanism or evidence requires it.
Practice 1: Equation or Function?
A company charges a fixed setup cost of \(\$200\) plus \(\$15\) per unit produced.
Choose a useful model structure and write the model.
Practice 2: Select a System
Two types of tickets are sold. A total of \(300\) tickets are sold for \(\$4{,}200\). Adult tickets cost \(\$18\) and student tickets cost \(\$10\).
Write a mathematical model.
Practice 3: Piecewise Structure
A service charges \(\$25\) for up to \(2\) hours and then \(\$8\) per additional hour.
Explain why a piecewise function is appropriate.
Practice 4: Recurrence or Closed Form?
A savings account is updated monthly by applying interest and then adding a fixed deposit.
Which representation is naturally closest to the mechanism?
Practice 5: Dimensional Check
Distance \(d\) is measured in miles, speed \(v\) in miles per hour, and time \(t\) in hours.
Which candidate is dimensionally consistent?
- \(d=v+t\)
- \(d=vt\)
Practice 6: Single Equation or System?
A factory must satisfy both a labor limit and a material limit while producing two products.
Which broad model structure is required?
Practice 7: Identify the Mechanism
A population grows by \(3\%\) each year under a constant-growth assumption.
What model family naturally matches this mechanism?
Practice 8: Algorithmic Model
An inventory system reorders whenever stock falls below a threshold.
Explain why an algorithm may be more natural than one unrestricted elementary formula.
Practice 9: Simplest Adequate Model
Over a narrow operating range, measured output rises by approximately \(4\) units whenever input rises by \(1\).
What simple model family should be considered first?
Practice 10: Hybrid Model
A delivery-planning model uses equations for travel time, capacity inequalities, a cost function, and a route-search procedure.
What kind of model is this?
Practice Answers
Practice 1
A linear cost function is natural:
\[ \boxed{C(q)=200+15q}. \]Practice 2
Let \(a\) be adult tickets and \(s\) student tickets.
\[ \boxed{ \begin{cases} a+s=300,\\ 18a+10s=4200. \end{cases}} \]Practice 3
The pricing rule changes after \(2\) hours, so different formulas apply in different regions. A piecewise function matches that threshold mechanism.
Practice 4
A recurrence or iterative algorithm is naturally closest to the mechanism because each month's new balance depends on the previous month's balance, interest, and deposit.
Practice 5
The dimensionally consistent model is
\[ \boxed{d=vt} \]because
\[ \frac{\text{miles}}{\text{hour}}\cdot\text{hour} = \text{miles}. \]Practice 6
A system is required because both resource limits must be represented simultaneously.
Practice 7
Constant percentage growth suggests an exponential model.
Practice 8
The system contains conditional logic: observe inventory, compare it with a threshold, decide whether to reorder, and then update the state. A step-by-step algorithm mirrors that process directly.
Practice 9
A linear model should be considered first because the observed rate of change is approximately constant:
\[ y\approx 4x+b. \]Practice 10
It is a hybrid model because it combines equations, inequalities, functions, and an algorithm.
What to Remember
- An equation expresses a condition that quantities must satisfy.
- A function emphasizes how outputs depend on inputs.
- A system represents several relationships or constraints simultaneously.
- An algorithm represents sequential, iterative, conditional, or computational processes.
- Many real models combine several structures.
- The mechanism should drive the choice of mathematics.
- Units, assumptions, constraints, and domains help screen candidate structures.
- The simplest adequate model is usually preferable to unnecessary complexity.
- A model can fit data without correctly representing the underlying mechanism.
The central habit is to choose the representation that makes the relationship, dependency, interaction, or process most explicit.
Where We Go Next
Once a mathematical structure has been selected, the next question is how much uncertainty the model should contain.
Some models produce one determined outcome from a given set of inputs. Others explicitly represent randomness, variation, or uncertain outcomes.
The next lesson, Deterministic and Stochastic Models, develops this distinction and shows how randomness changes both model structure and interpretation.