A residual is observed minus predicted; check the pattern and the modeling context.
What a strong response shows
Compute observed minus predicted.
Look for systematic residual patterns.
A small residual is not evidence of causation.
Loading the interactive investigation…
What You Need to Know
A model is useful when its form follows the data and its residuals show no systematic pattern.
Compare the model’s structure and residuals, not just one close prediction.
Fit candidate models, calculate residuals, and reject a model whose residuals curve or fan out.
Choosing the model with the smallest one-time error.
What does a curved residual pattern suggest?
Learning Objectives
2.6.A — Construct linear, quadratic, and exponential models based on a data set. 2.6.B — Validate a model constructed from a data set.
Match the prompt verb to the representation it asks you to build or interpret.
. Name the given quantities. . Choose the matching function or representation. . Show the parameter or operation that changes. . Check one value in context.
Starting calculations before identifying what the input and output mean.
Can you state what the input, output, and requested evidence are before choosing a formula?
Key Concepts
Linear models use constant differences, quadratic models use changing differences with stable second differences, and exponential models use ratios.
Use structure first, then use residuals or prediction error to validate.
A nearly constant second difference supports a quadratic model; a nearly constant ratio supports an exponential model.
Calling a model valid because it passes through two data points.
Which model is suggested by constant second differences?
Relevant Representations
Scatter plots, fitted curves, tables of residuals, and equations provide complementary model evidence.
Inspect the residual plot after inspecting the original fit.
| Candidate | Pattern to test | Warning sign |
|---|---|---|
| Linear | constant first difference | curved residuals |
| Quadratic | stable second difference | wrong turning behavior |
| Exponential | constant ratio | residual trend or negative prediction |
| Any model | small balanced residuals | systematic pattern |
Plot residuals against input. A good pattern is small, balanced, and without a visible curve or trend.
Reading a residual sign as a percent error without checking the units.
What should residuals look like for a well-matched model?
Essential Relationships or Formulas
Use an exact relationship to expose the pattern before substituting values.
Keep the variables and restrictions visible, then substitute with units attached.
Compute each residual from observed minus predicted, then look for a pattern.
Using a familiar formula without checking its domain, base, interval, or unit assumptions.
Which symbol or parameter carries the feature the prompt asks you to find?
Worked Examples
A short, labeled chain of steps is easier to check than a page of unexplained algebra.
Write the representation first, calculate second, and interpret last.
Fit the candidate models, compare residual patterns, and report the model that has both sensible structure and balanced residuals.
Skipping the interpretation step and reporting an answer with no units, domain, or model evidence.
Which step would you show so another student could reproduce the result?
Common Mistakes
Most errors come from a mismatch between the pattern, the operation, and the allowed input.
Before submitting, check pattern, operation, domain, and units in that order.
Circle the first line where the pattern or restriction changes. Correct that line, then recompute only the affected steps.
Trusting a calculator output that hides a wrong model, sign, base, or domain.
What independent check would catch your most likely error?
AP Exam Tips
Use the topic workflow to connect the representation, the equation, and the context before calculating.
Read the prompt, choose the smallest reliable representation, and verify the result in context.
. Identify the pattern. . Write the matching form. . Show the key substitution or transformation. . Check units, domain, and direction.
Submitting a number without explaining why the selected model or transformation fits.
Can you name the evidence that supports your final model or answer?
Open this topic’s practice set to apply the method. Work independently, then compare the evidence in your reasoning with the checked explanation.
Can you transfer the method?
A linear fit has small residuals at the first and last observations but a clear U-shaped residual pattern in between. Is endpoint agreement enough?
Compare your reasoning
No. Endpoint agreement misses systematic error in the middle. The curved residual pattern is evidence to investigate another model form and check its contextual meaning.
Work independently, then compare the mathematical evidence in your answer.