Skip to content
AP Statistics · Regression Analysis

Nonlinearity

Nonlinearity appears in 4 reviewed AP Statistics questions and sits in Regression Analysis. Below are worked examples showing exactly how the exam tests it, with the answer marked and every distractor explained.

How Nonlinearity is tested

Example 1Harder
A residual plot for a regression of y on x shows that residuals are large and positive for small x, near zero for medium x, and large and negative for large x. This pattern most likely indicates:
  1. AThe linear model is appropriate; residuals naturally decrease over the range of x.
  2. BThe observations are not independent of each other.
  3. CThe relationship between x and y is nonlinear.Correct
  4. DThere is increasing variance in y as x increases (heteroscedasticity).

Why C is right

A pattern where residuals go from large positive → near zero → large negative as x increases indicates that the true relationship bends downward more steeply than the linear model captures. The linear model over-predicts at high x-values and under-predicts at low x-values, suggesting a nonlinear (possibly curved or diminishing-returns) relationship.
Why the other 3 choices miss
A
Incorrect because naturally decreasing residuals over the range of x is precisely the type of pattern that indicates the linear model is NOT appropriate. A good residual plot would show random scatter.
B
Incorrect because independence violations appear as temporal or sequential patterns in residuals (e.g., autocorrelation), not as a consistent trend across x-values from positive to negative.
D
Incorrect because heteroscedasticity appears as a fan or funnel shape — increasing spread — in the residual plot. The described pattern (large positive → zero → large negative) is a directional trend, not a spread pattern.
Example 2Medium
A residual plot from a regression of crop yield on rainfall shows residuals that are randomly scattered around zero for low and moderate rainfall values but consistently negative for high rainfall values. What does this pattern suggest?
  1. AThere is no relationship between rainfall and crop yield at high rainfall levels
  2. BThe variance of the residuals increases with rainfall, violating the equal variance condition
  3. CThe linear model overpredicts crop yield at high rainfall levels, suggesting the relationship may curve downward at high rainfallCorrect
  4. DThe linear model underpredicts crop yield at all rainfall levels

Why C is right

Negative residuals mean observed values are below predicted values, so the model overpredicts at high rainfall. This suggests the true relationship curves downward at high rainfall (perhaps due to flooding or waterlogging), which a linear model cannot capture.
Why the other 3 choices miss
A
Incorrect because negative residuals indicate the model is still making predictions; the issue is that the predictions are systematically too high at high rainfall.
B
Incorrect because the pattern described is consistently negative residuals, not increasing spread; this indicates a linearity violation, not a variance violation.
D
Incorrect because the model does not underpredict at all levels — residuals are random at low and moderate rainfall and only negative at high rainfall.

Studied alongside

Other concepts from Regression Analysis.