Lesson 25: Worksheet
We are using two different kinds of “intervals” for very different kinds of purposes:
Confidence intervals are used to represent the precision of our estimates of coefficients. This always involves averaging over multiple data points.
Prediction intervals are used to indicate the range of likely values of the response variable when specifying the explanatory variables.
Almost always, such intervals are constructed at the “95% level.” Because of this we don’t always mention the level. But other “levels” can be used: 80%, 90%, 99%, and so on.
It’s hard to tell from a graph of a confidence interval (or band) what the confidence level is. On the other hand, it’s often straightforward to estimate the “level” for a prediction interval or band. For a 95% prediction level, for instance, about 5% of the data points will be outside the prediction interval, while for a 99% level only about 1% of the data points will be outside the prediction interval.
Here are graphs of some confidence bands and some prediction bands. For each graph, the sample size \(n\) is either 100 or 200. For each graph, say what the sample size is and whether it displays a prediction or a confidence band. If it’s a prediction band, estimate the prediction level.