Simulation

The diagram on the previous page effectively illustrates the increase in prediction variability when extrapolating. The diagram on this page also shows that the predictions are least variable at x-values close to the mean x-value in the data.

Click Accumulate, then take about 50 samples. The jittered dot plot on the right shows the errors when the mean response is estimated from the least squares lines.

The slider lets you see the distribution of these estimation errors at other x-values.

The predictions are least variable near x = 2.5, the mean x-value of all the data sets.

Click Show Theoretical Distn to see the normal distributions of these errors.