Function Modeling Tools for Polynomial and Rational Functions

Transformations of Functions

A transformation of a function is a systematic change to a function’s graph (and equation) that lets you build new models from familiar “parent” functions. This matters because in modeling you rarely start from scratch—you start from a basic shape (like a line, parabola, reciprocal curve, or cubic) and then shift, stretch, or reflect it to match a situation or a dataset.

The big idea is that changes outside the function affect outputs (vertical changes), while changes inside the function affect inputs (horizontal changes). Getting that “inside vs. outside” logic right is one of the most important skills in this unit.

Parent functions you’ll transform often

You’ll commonly use these as starting points:

• Linear: f(x) = x
• Quadratic: f(x) = x^2
• Cubic: f(x) = x^3
• Reciprocal (rational): f(x) = \frac{1}{x}

You may also see specific polynomials or rational functions treated as the “base” function being transformed.

The transformation template

A very useful general form is:

g(x) = a f(b(x-h)) + k

Here’s what each parameter does (conceptually):

• h: **horizontal shift**. The graph shifts right if h > 0 and left if h < 0.
• k: **vertical shift**. The graph shifts up if k > 0 and down if k < 0.
• a: **vertical scale** (stretch/compression) and possibly a **reflection** across the x-axis if a < 0.
• b: **horizontal scale** (stretch/compression) and possibly a **reflection** across the y-axis if b < 0.

Why the inside/outside distinction matters: x is the input. If you change the input before it enters f (the “inside”), you are changing where the graph’s features occur on the x-axis.

Vertical transformations (outside the function)

If you define:

g(x) = f(x) + k

then you add k to every output, moving the graph up or down.

If you define:

g(x) = a f(x)

then you multiply every output by a:

• If |a| > 1, the graph is stretched vertically (farther from the x-axis).
• If 0 < |a| < 1, it is compressed vertically.
• If a < 0, it is reflected across the x-axis.

Horizontal transformations (inside the function)

Horizontal effects often feel “backwards” because you’re modifying the input.

A shift:

g(x) = f(x-h)

moves the graph right by h (for h > 0). You can sanity-check: to get the same output you used to get at x = 0, you now need x = h because x-h = 0.

A horizontal scale:

g(x) = f(bx)

• If |b| > 1, the graph is compressed horizontally.
• If 0 < |b| < 1, it is stretched horizontally.
• If b < 0, it reflects across the y-axis.

Many students misremember this because vertical scaling multiplies outputs, but horizontal scaling multiplies inputs. A good way to remember: “Inside changes happen to x first, so they affect the horizontal placement.”

Working with key features (a faster, more reliable method)

Rather than plotting lots of points, you can transform key features:

• Intercepts
• Vertices/turning points
• Asymptotes (for rational functions)
• End behavior (what happens as x becomes very large positive or very large negative)

For a rational parent like f(x) = \frac{1}{x}, transformations move asymptotes:

g(x) = \frac{1}{x-h} + k

has vertical asymptote x = h and horizontal asymptote y = k.

Example 1: Transform a quadratic

Start with f(x) = x^2. Define:

g(x) = -2(x-3)^2 + 5

Interpretation:

• x-3 shifts right 3.
• The factor -2 reflects over the x-axis and stretches vertically by 2.
• +5 shifts up 5.

Key feature: the vertex of f(x) is at (0,0). After shifting right 3 and up 5, the vertex becomes (3,5). Reflection and vertical stretch change the “opening” and steepness, but the vertex location stays (3,5) because the vertex is anchored by the inside shift and outside shift.

Example 2: Transform a reciprocal function and identify asymptotes

Let f(x) = \frac{1}{x} and

g(x) = \frac{3}{x+2} - 1

Rewrite as:

g(x) = 3\cdot \frac{1}{x-(-2)} - 1

So:

• Vertical asymptote moves from x = 0 to x = -2.
• Horizontal asymptote moves from y = 0 to y = -1.
• Outputs are stretched by factor 3.

This feature-based approach is usually faster than plotting points.

Exam Focus

• Typical question patterns:
  • Given f and a graph or equation of g, describe the transformations that take f to g.
  • Identify key features (vertex, intercepts, asymptotes) after a transformation.
  • Match an equation in transformation form to its graph.
• Common mistakes:
  • Treating f(x-h) as a left shift instead of a right shift; fix this by testing where the original feature (like the vertex) must move.
  • Confusing horizontal scaling: f(2x) compresses horizontally, it does not stretch.
  • Forgetting that reflections occur when the multiplying factor is negative (outside for x-axis, inside for y-axis).

Function Model Selection and Construction

A function model is an equation (or rule) that represents a relationship between two quantities: an input (independent variable) and an output (dependent variable). Model selection means choosing a function family whose shape and features match the situation, and construction means determining specific parameters so the model fits given information.

This matters because Unit 1 focuses on polynomial and rational functions, which are common for modeling: polynomials can capture smooth growth and turning points, and rational functions can capture situations with division, rates, and asymptotic behavior (like “approaching a maximum,” “blowing up,” or having excluded values).

How to choose a model family (the “shape + context” method)

When you’re given a context, you usually decide based on:

1. What does the graph look like?
   • Straight trend: linear
   • One turning point: quadratic
   • Two turning points: cubic or higher-degree polynomial
   • Vertical asymptote or “undefined at a value”: rational
2. What do the inputs/outputs mean?
   • If the output is a ratio like “per unit,” rational models are common.
   • If you’re modeling area from a length, a quadratic might appear.
3. What are the key features you must match?
   • Intercepts (zeros)
   • End behavior
   • Maximum/minimum or turning points
   • Asymptotes and holes (rational)

Building polynomial models from zeros and behavior

If you know the zeros (the x-intercepts), you can build a polynomial using factors. If r is a zero, then x-r is a factor.

For example, a polynomial with zeros at x = -1 and x = 3 could start as:

P(x) = a(x+1)(x-3)

The constant a controls vertical scaling and (if negative) reflection.

If you also know a point on the graph, you can solve for a.

Worked example: Construct a quadratic with zeros at x = 1 and x = 5 and passing through (0,10).

Start with factor form:

P(x) = a(x-1)(x-5)

Use the point (0,10):

10 = a(0-1)(0-5)

10 = a(5)

a = 2

So the model is:

P(x) = 2(x-1)(x-5)

Why this works: zeros give you the structure; a single additional point fixes the scale.

Building rational models from asymptotes and intercepts

A simple transformed reciprocal model has the form:

R(x) = \frac{a}{x-h} + k

This is powerful because h and k directly tell you where the asymptotes are. But many rational functions in Unit 1 are more general:

R(x) = \frac{P(x)}{Q(x)}

where P and Q are polynomials and Q(x) \neq 0.

When constructing rational models, look for:

• Vertical asymptotes: values where Q(x) = 0 but the factor does not cancel.
• Holes: values where a factor cancels between numerator and denominator.
• Horizontal asymptotes (for many AP Precalculus situations): often determined by comparing degrees of numerator and denominator (a key structural idea for rational functions).

Worked example: Construct a rational function with vertical asymptote x = 2, horizontal asymptote y = -1, and passing through (3,1).

Use transformed reciprocal form:

R(x) = \frac{a}{x-2} - 1

Plug in (3,1):

1 = \frac{a}{3-2} - 1

1 = a - 1

a = 2

So:

R(x) = \frac{2}{x-2} - 1

Using transformations to construct models

Transformations aren’t just a graphing skill; they’re a modeling strategy. If you recognize that a situation “looks like” a parent function but shifted or scaled, you can build a model quickly and interpret parameters meaningfully.

For example, if a quantity approaches a limiting value over time and has a “blow-up” near a certain input, a rational transformation like \frac{a}{x-h} + k is a natural candidate.

Exam Focus

• Typical question patterns:
  • Choose an appropriate model type (polynomial vs. rational) given a context or graph, and justify the choice using features.
  • Construct a polynomial from given zeros and a point, or from a vertex and another point.
  • Construct a rational model from asymptotes and a point, then interpret parameters.
• Common mistakes:
  • Confusing a hole with a vertical asymptote; if a factor cancels, it’s a hole (removable discontinuity), not an asymptote.
  • Forgetting the scaling constant a when building from intercepts; intercepts alone don’t fix vertical stretch.
  • Ignoring domain restrictions implied by context (like time cannot be negative).

Piecewise-Defined Functions

A piecewise-defined function is a function defined by different formulas on different parts of its domain. You use piecewise definitions when a real situation has “rules that change,” like tax brackets, shipping costs with thresholds, or a physical model that behaves differently before and after a certain point.

Piecewise functions matter in this unit because polynomial and rational expressions often model different regimes: one expression might fit well for small inputs, while another fits for larger inputs, or there may be a structural break where behavior changes.

How piecewise functions work

A piecewise function looks like this:

f(x) = \begin{cases} \text{expression 1} & \text{condition 1} \\ \text{expression 2} & \text{condition 2} \end{cases}

However, to follow strict formatting rules here (no line breaks inside a single LaTeX block), it’s safer to present piecewise definitions in separate equations or in plain text format. For example:

Define f(x) by:

f(x) = x^2

for x < 1, and

f(x) = 2x+1

for x \ge 1.

The key idea is that for any allowed input, exactly one rule applies.

Continuity and “connecting” pieces

A common task is checking whether the graph has a jump or connects smoothly at a boundary value (like x = 1). At a boundary x = c you typically check:

• The left-hand value (using the rule for x < c)
• The right-hand value (using the rule for x > c)
• The actual defined value at x = c (if included)

If the pieces meet at the same point, the function is continuous at that boundary.

Worked example: Consider:

f(x) = x+2

for x < 0, and

f(x) = x^2

for x \ge 0.

Check continuity at x = 0:

Left value at 0 (using x+2):

f(0) = 0 + 2 = 2

Right value at 0 (using x^2):

f(0) = 0^2 = 0

They do not match, so there is a jump discontinuity at x = 0.

Piecewise models in context

In modeling, boundaries often come from:

• A physical limit (after a machine reaches capacity)
• A policy change (pricing changes after a certain number of units)
• Different formulas in different regimes (approximation methods)

When interpreting piecewise functions, always state what each interval means in words. That’s usually what turns a correct algebraic answer into a correct modeling answer.

Graphing piecewise functions (and endpoint notation)

When graphing:

• Use an open circle at an endpoint if the interval is strict (like x < 2).
• Use a closed circle if the endpoint is included (like x \le 2).

A frequent error is drawing the correct curve but marking the wrong type of endpoint—this changes the function’s actual value at the boundary.

Exam Focus

• Typical question patterns:
  • Evaluate a piecewise function at a value and justify which rule applies.
  • Determine whether a piecewise function is continuous at a boundary.
  • Write a piecewise function to model a real-world scenario (cost, rate, constraints).
• Common mistakes:
  • Using the wrong piece because you ignore the interval condition; always check the inequality first.
  • Marking open vs. closed circles incorrectly when graphing.
  • Assuming continuity automatically; you must compare the boundary values from both sides.

Composition of Functions

The composition of functions combines two functions so that the output of one becomes the input of the other. If you have functions f and g, the composition f \circ g is defined by:

(f \circ g)(x) = f(g(x))

This matters for modeling because many real processes happen in stages. For example, you might convert units (first function) and then apply a cost formula (second function), or compute a radius from a volume and then compute surface area from the radius.

How composition works (the input-output machine view)

Think of g as the first machine: you input x and get g(x). Then f is the second machine: you input g(x) and get f(g(x)).

Order matters. In general:

f(g(x)) \neq g(f(x))

So you must read carefully which function is “inside.”

Domain of a composition

A composition is only defined when:

1. x is in the domain of g, and
2. g(x) is in the domain of f.

This becomes especially important with rational functions, because certain inputs may make denominators zero.

Worked example (including domain):

Let:

f(x) = \frac{1}{x-1}

and:

g(x) = x^2 - 4

Find (f \circ g)(x) and its domain.

Compose:

(f \circ g)(x) = f(g(x)) = \frac{1}{(x^2 - 4) - 1}

So:

(f \circ g)(x) = \frac{1}{x^2 - 5}

Domain restriction comes from denominator nonzero:

x^2 - 5 \neq 0

So:

x \neq \sqrt{5}

and:

x \neq -\sqrt{5}

Composition as model building

Composition lets you build more realistic models from simpler ones:

• A polynomial might model production as a function of time, P(t).
• A rational function might model cost per unit as a function of production, C(p).
• Then C(P(t)) models cost per unit as a function of time.

The key modeling skill is interpreting what the inside function represents and what the outside function does to that quantity.

Decomposing a function (reverse thinking)

Sometimes you’re asked to express a function as a composition. For example:

h(x) = \sqrt{x+3}

You can think of this as “add 3, then take the square root,” meaning:

• g(x) = x+3
• f(x) = \sqrt{x}
• h(x) = f(g(x))

Even when square roots aren’t the focus of Unit 1, this decomposition skill carries over directly to polynomial and rational forms.

Exam Focus

• Typical question patterns:
  • Compute f(g(x)) or g(f(x)) from given formulas.
  • Determine the domain of a composition, especially with rational functions.
  • Interpret a composition in context (what quantity is being fed into what relationship).
• Common mistakes:
  • Reversing the order: f \circ g means apply g first.
  • Ignoring domain restrictions introduced by the inside output (for instance, when g(x) = 1 makes the outer denominator zero).
  • Simplifying incorrectly after substitution (a common algebra slip with negatives or parentheses).

Regression and Residual Analysis

In many modeling problems, you don’t get an exact formula—you get data. Regression is the process of finding a function that best fits a set of data points. In AP Precalculus, regression is typically done with technology (graphing calculator or software), but you are responsible for interpreting the results, selecting an appropriate model type, and evaluating how good the fit is.

A residual measures the vertical error between the observed data value and the model’s predicted value. Residual analysis is how you check whether your chosen model is appropriate.

What regression is doing (conceptually)

Suppose you have data points (x_i, y_i) and a model \hat{y} = f(x). For each point, the model predicts:

\hat{y}_i = f(x_i)

The residual is:

r_i = y_i - \hat{y}_i

• If r_i > 0, the model underestimates (point is above the curve).
• If r_i < 0, the model overestimates.

Residuals matter because a model can look decent on the main plot but still have a systematic bias that shows up clearly in residuals.

Choosing a regression type in this unit

Within Unit 1’s polynomial/rational focus, common regression choices include:

• Linear regression when the trend is approximately straight.
• Quadratic (or higher polynomial) regression when the data show curvature and potentially turning behavior.

Rational regression can exist as a concept, but on many platforms it’s less standard than polynomial regression; what you can always do is reason structurally: if the data suggest an asymptote or blow-up behavior, a rational model may be more appropriate than a polynomial.

A practical selection strategy:

1. Make a scatter plot.
2. Try a candidate model (like linear or quadratic regression).
3. Check residuals for randomness vs. pattern.
4. Prefer simpler models if fits are comparable (a modeling principle, not just a math trick).

Reading and using a regression equation

If your technology outputs a quadratic regression:

\hat{y} = ax^2 + bx + c

Interpretation is not just “those are the coefficients.” You should connect:

• The sign of a to opening direction (up if a > 0, down if a < 0)
• The size of |a| to curvature (bigger magnitude means tighter curve)
• The intercept c to predicted value when x = 0 (if x = 0 is meaningful)

Always check whether extrapolating beyond the data’s x-range makes sense; polynomial models can behave wildly outside the observed interval.

Residual plots: what you want to see

A residual plot graphs x_i on the horizontal axis and residuals r_i on the vertical axis.

A good sign: residuals scattered around 0 with no clear pattern.

A warning sign: a clear curve, trend, or clustering above/below 0. That suggests the model is missing structure (for example, using a line when the relationship is curved).

Example: Compute and interpret residuals

Suppose your model is:

\hat{y} = 2x + 1

and you have a data point (3, 8).

Compute prediction:

\hat{y} = 2(3) + 1 = 7

Residual:

r = 8 - 7 = 1

Interpretation: the actual value is 1 unit above the model’s prediction at x = 3.

Example: Using residual patterns to improve a model

Imagine you fit a linear regression and the residual plot shows negative residuals for small x, positive residuals for middle x, and negative residuals again for large x (a “smile” pattern). That pattern indicates the data curve upward and then back relative to the line—often suggesting a quadratic model would capture the curvature better.

This kind of reasoning is what residual analysis is for: it tells you not just “the model is imperfect” (all models are), but “the model is missing a specific kind of shape.”

Regression vs. exact models

In earlier sections, you built functions to match exact features (like exact zeros or asymptotes). Regression is different: it optimizes fit across many points but usually does not hit any point exactly.

A common misunderstanding is thinking regression “finds the true function.” It finds a best-fit function within the chosen family. If you choose the wrong family, you can get a misleading equation even if the calculator gives you one.

Exam Focus

• Typical question patterns:
  • Given data and multiple candidate models, select the best model using scatter plots and residual plots.
  • Interpret residuals (sign and magnitude) and what they say about prediction error.
  • Explain, in context, why a linear model is insufficient and a polynomial model is better (or vice versa).
• Common mistakes:
  • Using r_i = \hat{y}_i - y_i instead of r_i = y_i - \hat{y}_i; always remember “observed minus predicted.”
  • Declaring a model good just because it looks close on the main graph, without checking residual patterns.
  • Extrapolating far beyond the data range with a polynomial regression and treating it as reliable.