BetterGrades Precalculus · Unit 16 · Lesson
Model selection and comparison
Select and compare function models using patterns, residuals, domains, and parameter meanings.
The problem that opens the lesson
Three models fit the same ten data points: linear, exponential, and logistic. What evidence should decide among them?
Solution
Begin by identifying the mathematical object and the information that fixes it. Define the decision question, fit or derive candidates, inspect residuals, interpret parameters, test domain and limiting behavior, and defend one model with stated limitations. The relevant conditions are not optional bookkeeping: A more complicated model can overfit noise. A simpler model can miss real structure. Following that structure gives Residual patterns, parameter plausibility, domain, long-run behavior, and mechanism.
Why this works
Two models can fit the observed interval similarly while producing radically different extrapolations. The calculation and the representation should agree, so a graph, diagram, table, or substitution check should support the same conclusion.
What this lesson is really about
Model selection compares how well different function families explain data, structure, and context.
Residual size matters, but so do residual pattern, parameter meaning, domain behavior, long-run prediction, simplicity, and mechanism.
The point is not merely to reproduce a formula. A learner should be able to identify the quantities or geometric objects involved, explain why the relationship has its stated form, and recognize when the same idea appears in a graph, table, diagram, or model.
Why the relationship works
Two models can fit the observed interval similarly while producing radically different extrapolations.
A reliable way to work
Define the decision question, fit or derive candidates, inspect residuals, interpret parameters, test domain and limiting behavior, and defend one model with stated limitations.
A more complicated model can overfit noise. A simpler model can miss real structure.
After the symbolic work is complete, check the result. Depending on the lesson, this may mean substituting into an original equation, comparing coordinates, examining a graph, checking units, testing an interval, or confirming that every branch of a periodic solution has been included.
What commonly goes wrong
A common error is choosing the highest fit statistic without evaluating plausibility or uncertainty.
The repair is to return to the definition and identify the first step where the invalid solution stops describing the original mathematical object. Later algebra cannot rescue a first step that changed the domain, orientation, branch, or meaning of the problem.
Worked examples
Worked example 1
Three models fit the same ten data points: linear, exponential, and logistic. What evidence should decide among them?
Solution
Begin by identifying the mathematical object and the information that fixes it. Define the decision question, fit or derive candidates, inspect residuals, interpret parameters, test domain and limiting behavior, and defend one model with stated limitations. The relevant conditions are not optional bookkeeping: A more complicated model can overfit noise. A simpler model can miss real structure. Following that structure gives Residual patterns, parameter plausibility, domain, long-run behavior, and mechanism.
Why this works
Two models can fit the observed interval similarly while producing radically different extrapolations. The calculation and the representation should agree, so a graph, diagram, table, or substitution check should support the same conclusion.
Transfer example
Problem
Compare sinusoidal and polynomial seasonal fits.
Worked development
Define the decision question, fit or derive candidates, inspect residuals, interpret parameters, test domain and limiting behavior, and defend one model with stated limitations. In this example, the first useful move is to make the defining structure visible rather than to search for a memorized answer. Residual size matters, but so do residual pattern, parameter meaning, domain behavior, long-run prediction, simplicity, and mechanism. Then apply the conditions explicitly: A more complicated model can overfit noise. A simpler model can miss real structure. Finish by checking the result in a second representation and explaining what the result means.
Interpretation
Model comparison supports scientific reasoning, economics, engineering, and data analysis.
Reasoning example
Problem
Reject a rational model with a contextual asymptote.
Worked development
Define the decision question, fit or derive candidates, inspect residuals, interpret parameters, test domain and limiting behavior, and defend one model with stated limitations. In this example, the first useful move is to make the defining structure visible rather than to search for a memorized answer. Residual size matters, but so do residual pattern, parameter meaning, domain behavior, long-run prediction, simplicity, and mechanism. Then apply the conditions explicitly: A more complicated model can overfit noise. A simpler model can miss real structure. Finish by checking the result in a second representation and explaining what the result means.
Interpretation
Model comparison supports scientific reasoning, economics, engineering, and data analysis.
Worked example 4: quick check
Name two reasons to reject a numerically good fit.
Solution
Begin by identifying the mathematical object and the information that fixes it. Define the decision question, fit or derive candidates, inspect residuals, interpret parameters, test domain and limiting behavior, and defend one model with stated limitations. The relevant conditions are not optional bookkeeping: A more complicated model can overfit noise. A simpler model can miss real structure. Following that structure gives Implausible parameters, wrong domain behavior, systematic residuals, or unsupported extrapolation.
Why this works
Two models can fit the observed interval similarly while producing radically different extrapolations. The calculation and the representation should agree, so a graph, diagram, table, or substitution check should support the same conclusion.
Read this graph as text
Model selection and comparison · Competing fits and extrapolation. Follow the foundation example from its given information to the conclusion. The labels identify the mathematical feature that makes the result valid: Two models can fit the observed interval similarly while producing radically different extrapolations. The calculation and the representation should agree, so a graph, diagram, table, or substitution check should support the same conclusion. The figure uses concrete points, curves, arrows, intervals, or matrix structure instead of relying on color alone.
Labels, point shapes, line styles, arrows, and position carry the mathematical meaning; color is supplementary.
Why it matters: Use the mathematical objects in this figure to support the lesson outcome: Select and compare function models using patterns, residuals, domains, and parameter meanings.
Follow the foundation example from its given information to the conclusion. The labels identify the mathematical feature that makes the result valid: Two models can fit the observed interval similarly while producing radically different extrapolations. The calculation and the representation should agree, so a graph, diagram, table, or substitution check should support the same conclusion.
Read this graph as text
Model selection and comparison · Residual-pattern comparison. Read the numbered reasoning path in order. Each stage preserves the quantities, restrictions, or structural conditions needed for model selection and comparison. The figure uses concrete points, curves, arrows, intervals, or matrix structure instead of relying on color alone.
Labels, point shapes, line styles, arrows, and position carry the mathematical meaning; color is supplementary.
Why it matters: Use the mathematical objects in this figure to support the lesson outcome: Select and compare function models using patterns, residuals, domains, and parameter meanings.
Read the numbered reasoning path in order. Each stage preserves the quantities, restrictions, or structural conditions needed for model selection and comparison.
Read this graph as text
Model selection and comparison · Parameter plausibility checklist. Compare the valid path with the tempting shortcut. The figure shows why choosing the highest fit statistic without evaluating plausibility or uncertainty leads to a false conclusion. The figure uses concrete points, curves, arrows, intervals, or matrix structure instead of relying on color alone.
Labels, point shapes, line styles, arrows, and position carry the mathematical meaning; color is supplementary.
Why it matters: Use the mathematical objects in this figure to support the lesson outcome: Select and compare function models using patterns, residuals, domains, and parameter meanings.
Compare the valid path with the tempting shortcut. The figure shows why choosing the highest fit statistic without evaluating plausibility or uncertainty leads to a false conclusion.
Application and interpretation
Model comparison supports scientific reasoning, economics, engineering, and data analysis.
A contextual answer must include units, a meaningful domain, and the assumptions that make the model plausible. An exact mathematical relationship should not be diluted into a decimal unless a measurement or comparison requires it.
Name two reasons to reject a numerically good fit.
Write a complete attempt before opening the exact answer.
Attempt once to unlock the answer
Complete a substantive attempt before revealing the server-held answer.
Ten concrete questions
01Name two reasons to reject a numerically good fit.
Write a complete attempt before opening the exact answer.
Attempt once to unlock the answer
Complete a substantive attempt before revealing the server-held answer.
02Compare sinusoidal and polynomial seasonal fits.
Write a complete attempt before opening the exact answer.
Attempt once to unlock the answer
Complete a substantive attempt before revealing the server-held answer.
03Reject a rational model with a contextual asymptote.
Write a complete attempt before opening the exact answer.
Attempt once to unlock the answer
Complete a substantive attempt before revealing the server-held answer.
04Explain why lowest residual sum alone may be insufficient.
Write a complete attempt before opening the exact answer.
Attempt once to unlock the answer
Complete a substantive attempt before revealing the server-held answer.
05State the defining idea behind model selection and comparison in one precise sentence.
Write a complete attempt before opening the exact answer.
Attempt once to unlock the answer
Complete a substantive attempt before revealing the server-held answer.
06What condition or domain restriction must remain visible in the solution?
Write a complete attempt before opening the exact answer.
Attempt once to unlock the answer
Complete a substantive attempt before revealing the server-held answer.
07Describe the most likely incorrect first step and explain why it fails.
Write a complete attempt before opening the exact answer.
Attempt once to unlock the answer
Complete a substantive attempt before revealing the server-held answer.
08Translate the main result into a second representation: graph, diagram, table, equation, or context.
Write a complete attempt before opening the exact answer.
Attempt once to unlock the answer
Complete a substantive attempt before revealing the server-held answer.
09Write one exact conclusion and one corresponding numerical approximation or verbal interpretation.
Write a complete attempt before opening the exact answer.
Attempt once to unlock the answer
Complete a substantive attempt before revealing the server-held answer.
10Explain how this lesson's idea will be used later in the course.
Write a complete attempt before opening the exact answer.
Attempt once to unlock the answer
Complete a substantive attempt before revealing the server-held answer.
Lesson summary
Model selection compares how well different function families explain data, structure, and context.
The central condition to remember is this: A more complicated model can overfit noise. A simpler model can miss real structure.
Connection forward
The next lesson revisits average rate of change across function families.
The next lesson is Average rate of change revisited.
Source record
Original BetterGrades manuscript, rights-separated references.
- AP Precalculus mathematical practices
- Lippman & Rasmussen, rates of change and function behavior
- BetterGrades Calculus Limits and Continuity course
- Stitz & Zeager, function synthesis and numerical methods
No long source passage is reproduced.