1

Plot with CI and PI

I have this graph that depicts PI being larger than CI for values of 14.5 and 24 for the independent/predictor variable. CI take into account regression coefficients, which are estimates. The PI has true error, which I am not sure what that means or where that comes from. Is that the noise captured in epsilon? Why is the PI larger than the CI?

Matthew
  • 77
  • 6
  • 1
    It is larger because you are trying to predict where a new observation will be & that new observation has an additional standard deviation of the error term. Confidence intervals are based on only the fitted values & do not involve making a prediction. It represents the uncertainty in the "fitted' value. – Michael R. Chernick Oct 09 '19 at 02:37
  • @Matthew To the two components of the confidence interval [discussed here](https://stats.stackexchange.com/questions/85560/shape-of-confidence-interval-for-predicted-values-in-linear-regression/85565#85565) you have to add a term for the variation in the data about the population line. – Glen_b Oct 09 '19 at 04:40

0 Answers0