Wize Business Analytics Textbook > Correlation and Regression

Prediction Intervals for a New Observation

Prediction intervals for individual outcomes

A fitted regression line estimates the average response at a chosen value of xx, but an individual future observation can lie above or below that average. A prediction interval accounts for both uncertainty in the fitted line and the ordinary person-to-person or measurement-to-measurement scatter around it.

A prediction interval includes the scatter of one future response as well as uncertainty in the fitted mean.

A prediction interval includes the scatter of one future response as well as uncertainty in the fitted mean.


For a new response at x0x_0, start with the fitted value y^0\hat y_0. A two-sided prediction interval has the form

y^0±t∗s1+1n+(x0−xˉ)2Sxx.\hat y_0 \pm t^*s\sqrt{1+\frac{1}{n}+\frac{(x_0-\bar x)^2}{S_{xx}}}.

Here ss is the residual standard error, nn is the sample size, and Sxx=∑(xi−xˉ)2S_{xx}=\sum(x_i-\bar x)^2. The leading 11 is crucial: a new individual has its own unpredictable residual.


That is why a prediction interval is wider than an interval for the mean response at the same x0x_0.

Relative prediction-interval width across the predictor range

Relative prediction-interval width across the predictor range

Worked example

Suppose a laboratory has y^0=31.6\hat y_0=31.6, t∗=2.05t^*=2.05, s=2.8s=2.8, n=16n=16, and the square-root factor is 1.111.11. The margin is 2.05(2.8)(1.11)2.05(2.8)(1.11), so add and subtract that margin from the fitted value. The resulting interval describes where one newly measured outcome is expected to fall, assuming the regression conditions are reasonable.


As the figure suggests, intervals also widen away from xˉ\bar x, because the line is less precisely anchored at extreme predictor values. Avoid calling this a probability statement about the one outcome already observed.


Before data are collected, the procedure captures the next response at its stated confidence level; after it is observed, it either falls in the interval or it does not.


Common mistake: dropping the leading 11. That produces the narrower interval for an average response, not the range needed for one new observation.

In the figure, what does the red vertical span at x0x_0 represent?

A prediction interval includes the scatter of one future response as well as uncertainty in the fitted mean.

A prediction interval includes the scatter of one future response as well as uncertainty in the fitted mean.

Relative prediction-interval width across the predictor range

Relative prediction-interval width across the predictor range

Choose an answer

Create a free account to get started

Unlock full course notes by signing up

  • Video lessons taught by top educators
  • Practice questions with guided solutions
  • Written review notes for reference
  • Expert answers anytime from course tutors

Which term makes the standard-error factor for a prediction interval for one new response larger than the corresponding factor for a mean response?

Choose an answer

Create a free account to get started

Unlock full course notes by signing up

  • Video lessons taught by top educators
  • Practice questions with guided solutions
  • Written review notes for reference
  • Expert answers anytime from course tutors

At the same x0x_0, an instructor reports a 95% interval for the mean exam mark and a 95% prediction interval for one new student's mark. Which statement is correct?

Choose an answer

Create a free account to get started

Unlock full course notes by signing up

  • Video lessons taught by top educators
  • Practice questions with guided solutions
  • Written review notes for reference
  • Expert answers anytime from course tutors