The shape of a relationship is only known around the data

The models that we have used to describe the relationship between a response, Y, and explanatory variable, X, are usually only approximations to the 'real' relationship. For example, a scatterplot may look linear, but we really have no information about the shape of the relationship beyond our data.

A model may be useful for predicting Y from values of X that are within the range of x-values in our data, but we should be very cautious about using it to predict Y outside this range. This is called extrapolation and it can be badly in error.

Avoid using a model to predict Y far beyond the available data.