PrecalcIQ · AP Precalculus · Lesson 9 of 25
PrecalcIQ · AP Precalculus

Lesson 09: Function Model Selection, Construction & Application

Unit 1 · Phase 1

Objectives

Warm-Up

Four one-line stories. Name the function family:

  1. "Every month the subscription adds 40 users." (constant rate → linear)
  2. "Revenue climbs, peaks in July, falls symmetrically." (one extremum → quadratic)
  3. "The river's depth rises, dips, rises again along its course." (multiple extrema → higher-degree polynomial)
  4. "Average cost per unit keeps falling but can never drop below $2." (horizontal asymptote → rational)

Model selection is translation: behavior described in words → feature owned by a family. The exam gives the story; you supply the family and defend the choice with rates of change.


Core Concept

Selecting a model from data

With equally spaced inputs, run the diagnostics from Lesson 2:

Pattern in outputs Model
constant first differences linear
constant second differences quadratic
constant third differences cubic
roughly constant ratios exponential (Unit 2)
approaches a fixed value rational or exponential (asymptote)

Real data are noisy — the honest claim is "approximately constant second differences, so a quadratic model is appropriate." (Hedged language earns points; certainty about noisy data loses them.)

Selecting a model from context

Structural clues in the story:

Constructing the model

Three construction routes, by what you're given:

  1. Key features → transformation or factored form: vertex (h, k) → a(x − h)² + k; zeros → a(x − r₁)(x − r₂)…; then one more point pins a.
  2. Exact data → solve for coefficients (second differences give the quadratic's a; systems of equations for the rest).
  3. Noisy data → 📱 regression (STAT → CALC → LinReg/QuadReg/CubicReg on TI; Statistics app elsewhere). Report coefficients to three decimal places and store the full-precision model for later parts — rounding early then computing is a classic point-loser.

Applying the model (where the FRQ points live)

[GRAPH: Scatterplot of rise-then-fall data points (days 1–9 vs. attendance) with a concave-down parabola fit through them; the parabola extended as a dashed curve past day 13 dipping below the x-axis, region labeled "model breakdown — negative attendance".]


Worked Examples

Example 1 (easy) — Family from context 🚫 No-Calc

Problem: The average cost per phone of producing x phones is A(x) = (50,000 + 30x)/x. Why is a rational model structurally right here, and what does its horizontal asymptote mean?

Solution: Average cost is total cost ÷ quantity — a ratio of two functions of x, which is what a rational function is. Degrees are equal (1 = 1), so lim x→∞ A(x) = 30/1 = 30: as production grows, average cost per phone approaches $30 (the per-unit cost, once the $50,000 fixed cost is spread thin). The asymptote is the floor: no production level makes the average dip below $30.

Interpretation: "Per," "average," and "concentration" are rational-function trigger words.

Example 2 (medium) — Construct from features 🚫 No-Calc

Problem: A fountain's water arc peaks 8 feet above the ground at a horizontal distance of 3 feet from the nozzle, and the water lands 7 feet from the nozzle. Model the arc's height h as a function of horizontal distance x.

Strategy: Vertex given → vertex form; landing point pins a.

Solution: Vertex (3, 8): h(x) = a(x − 3)² + 8. Landing at (7, 0):

0 = a(16) + 8  →  a = −1/2
h(x) = −(1/2)(x − 3)² + 8

Sanity: a < 0 ✓ (arc opens downward); nozzle height h(0) = −4.5 + 8 = 3.5 ft ✓ plausible.

Interpretation: Choose the form the givens hand you — a vertex means vertex form; don't fight your way there from standard form.

Example 3 (medium) — Exact data to model 🚫 No-Calc

Problem: Values: f(1) = 5, f(2) = 12, f(3) = 23, f(4) = 38. Find a model and predict f(6).

Solution: Δ = 7, 11, 15; Δ² = 4, 4 → quadratic with a = Δ²/(2h²) = 4/2 = 2. Then f(x) = 2x² + bx + c: f(1) = 2 + b + c = 5 and f(2) = 8 + 2b + c = 12 → subtract: 6 + b = 7 → b = 1, c = 2.

f(x) = 2x² + x + 2      f(6) = 72 + 6 + 2 = 80

Check f(4): 32 + 4 + 2 = 38 ✓

Interpretation: Exact data → algebra recovers the model precisely; regression is for noise.

Example 4 (AP-style) — Regression workflow 📱 Calc-Active

Problem: A biologist tracks an invasive plant's coverage (m²): (week, area) = (0, 2.1), (2, 4.8), (4, 12.2), (6, 23.9), (8, 40.1). Fit a quadratic model and estimate coverage in week 9.

Solution (workflow): 1. Enter weeks in L1, areas in L2. 2. QuadReg → A(t) ≈ 0.559t² + 0.284t + 2.071 (coefficients to three decimals; store the regression equation to Y1 unrounded). 3. Evaluate Y1(9) — using full precision — ≈ 49.9 m².

Write-up: "Using quadratic regression, A(t) ≈ 0.559t² + 0.284t + 2.071. A(9) ≈ 49.9, so the model predicts about 49.9 m² of coverage in week 9."

Interpretation: The graded moves: three-decimal coefficients, full-precision evaluation, contextual sentence with units. (Week 9 is just past the data's edge — near-extrapolation is acceptable with a caveat; week 30 would not be.)


Common Mistakes

  1. Choosing the family by vibes instead of rates. "It goes up fast → exponential!" Quadratics also go up fast. Fix: difference/ratio analysis on tables; named features (extrema, asymptotes) from contexts.
  2. Rounding the regression, then computing with the rounded model. Coefficients reported to 3 decimals are for the reader; computations use the stored full-precision equation. Error compounds fast when t² multiplies a truncated coefficient.
  3. Predictions without units or context. "f(9) = 50.4" is half credit at best. Say what the 50.4 is.
  4. Treating the model as true everywhere. Quadratic attendance goes negative eventually; average cost isn't defined at x = 0. State the breakdown, don't ignore it.
  5. Forcing a high-degree polynomial through every data point. A degree-7 curve through 8 points fits perfectly and predicts garbage. The exam wants the simplest family consistent with the pattern.

Practice Problems

Question 1
🚫 A town's daily high temperature rises through spring to a single midsummer peak, then falls symmetrically through autumn. Over the year, the most appropriate model family is
Question 2
🚫 Over equally spaced inputs, a data set's third differences are approximately constant (and nonzero). The appropriate model is
Question 3
🚫 Which quantity most naturally calls for a rational-function model?
Question 4
🚫 f is quadratic with f(1) = 5, f(2) = 12, f(3) = 23. The value of f(6) is
Question 5
🚫 A linear model for a library's collection is B(t) = 12t + 340 (books, t in months). The 12 means
Question 6
🚫 A graph of data shows exactly three local extrema. The minimum possible degree of a polynomial model is
Question 7
🚫 A quadratic model fit to sales data from weeks 1–10 predicts sales of −1,200 units in week 20. The best conclusion is
Question 8
🚫 Hot cocoa cools toward room temperature, 21°C, never reaching it. A model for its temperature should have
Question 9
🚫 A company's profit rises, peaks, then falls, and is zero at production levels 200 and 1,400. A quadratic model P(x) = a(x − 200)(x − 1400) requires
Question 10
📱 A ball's height is h(t) = −4.9t² + 24t + 1.2 meters. The FIRST time the ball reaches a height of 20 meters is approximately
Question 11
🚫 A model's input is time in minutes and output is volume in liters. The units of its average rate of change are

12. (FRQ-style) 📱 The table shows a lake's algae coverage:

week t 0 1 2 3 4
area (m²) 10 14.5 21.2 30.1 41.5

(i) Explain, using ratios or differences, why a quadratic model is more appropriate than a linear one. (ii) Use quadratic regression to find a model A(t); write coefficients to three decimal places. (iii) Use the model to predict coverage in week 6, and state one limitation of that prediction.


FRQ Practice — Task Model: Modeling a Non-Periodic Context (FRQ 1 style) 📱 Calc-Active

Daily attendance at a county fair, in hundreds of visitors, was recorded on five days:

day d 1 3 5 7 9
attendance N (hundreds) 36.5 60.5 72.5 72.5 60.5

(a) (i) Find the average rate of change of N over the interval [3, 7]. (ii) Interpret its meaning in context, with units.

(b) (i) Explain why a quadratic model is appropriate for these data, using rates of change over equal-length intervals. (ii) Find a quadratic regression model for N(d).

(c) (i) Use the model to predict attendance on day 11. (ii) The fair's organizers want to use the model to plan for day 14. Explain one reason the model should not be used for that prediction.

Model Response & Rubric (6 points)

(a) [2 pts] (i) [1 pt] (N(7) − N(3))/(7 − 3) = (72.5 − 60.5)/4 = 3 hundred visitors per day. (ii) [1 pt] From day 3 to day 7, daily attendance increased by an average of about 300 visitors per day. (Value + units + context.)

(b) [2 pts] (i) [1 pt] Over equal 2-day intervals, the rates of change are (60.5−36.5)/2 = 12, (72.5−60.5)/2 = 6, (72.5−72.5)/2 = 0, (60.5−72.5)/2 = −6 hundred visitors/day: the rate of change decreases by a constant 6 each interval. Rates changing at a constant rate is the signature of a quadratic. (ii) [1 pt] Quadratic regression: N(d) = −1.5d² + 18d + 20 (here the data fit exactly; on the real exam report a ≈ −1.500, b ≈ 18.000, c ≈ 20.000 and store the full-precision model).

(c) [2 pts] (i) [1 pt] N(11) = −1.5(121) + 18(11) + 20 = −181.5 + 198 + 20 = 36.5 → about 3,650 visitors on day 11. (ii) [1 pt] Any one: the model predicts N(14) = −294 + 252 + 20 = −22 (negative attendance — impossible, the model breaks down); day 14 is far outside the observed days 1–9, so extrapolation is unreliable; the fair may not even run 14 days. (One clearly stated limitation earns the point.)


Show answer key & explanations

(g) Answer Key

1. (C). Single peak with symmetric rise and fall → one extremum → quadratic. (D) would need more turning points; (B) needs asymptotic behavior.

2. (A). Constant third differences ↔ degree 3. (B) is second differences.

3. (D). Average cost = total ÷ quantity: a ratio → rational (Example 1). (A) linear; (B) quadratic; (C) exponential.

4. (A). From Example 3: f(x) = 2x² + x + 2, f(6) = 80. (B)/(D) arise from wrong b or c; (C) from a = 2.5.

5. (B). Slope of a linear model = constant rate: 12 books per month. (A) confuses slope with intercept (340).

6. (C). n extrema needs degree ≥ n + 1 → at least 4. (A) forgets the +1; (B)/(D) overshoot the minimum.

7. (D). Week 20 is double the data window; the impossible prediction signals extrapolation failure, not a computing error. (A) trusts the model over reality; (B)/(C) misdiagnose — the quadratic may fit weeks 1–10 perfectly well.

8. (B). "Approaches but never reaches 21" is a horizontal asymptote at y = 21. (A) would mean the cocoa hits 21°C when the variable equals... a zero is an output of 0; (C) confuses axes.

9. (A). Rise-peak-fall = concave down → a < 0. The zeros alone don't fix the sign (that's the trap in (D)) — but the described shape does.

10. (B). Solve −4.9t² + 24t + 1.2 = 20 → 4.9t² − 24t + 18.8 = 0 → t = (24 ± √(576 − 368.48))/9.8 = (24 ± √207.52)/9.8 → t ≈ 0.979 or 3.919. First time: 0.979 s. (A) is the second crossing (on the way down); (C) is the vertex time (24/9.8); (D) halves the correct answer.

11. (D). Δoutput/Δinput = liters per minute. (B) inverts; (C) multiplies.

12. (FRQ-style, 6 points) (i) [2 pts] First differences: 4.5, 6.7, 8.9, 11.4 — not constant, so linear is inappropriate; second differences: 2.2, 2.2, 2.5 — approximately constant, consistent with a quadratic. (Hedged language for noisy data.) (ii) [2 pts] QuadReg: A(t) ≈ 1.143t² + 3.289t + 10.026 (three decimals; store full precision). (Any correctly executed regression on these data earning coefficients within rounding of these values receives credit.) (iii) [2 pts] A(6) ≈ 70.9 m² (using stored precision). Limitation: week 6 lies outside the observed weeks 0–4, so this extrapolation assumes the growth pattern continues unchanged — algae growth could saturate (the lake is finite), making the quadratic unreliable beyond the data.


🎯 Exam tip: FRQ 1 has a rhythm: compute → interpret → model → predict → doubt. The first four are mechanical if you've practiced the calculator workflow. The fifth — articulating a limitation — is free points for anyone who memorizes the two magic sentences: "outside the observed data" and "the model predicts impossible values beyond x = …".

← All lessons
Lesson 10 ›
Score: 0/0 correct