Randomness
We usually need to acknowledge that the data we have collected are random — the values could easily have been different. We need a way to generalise from the data to the underlying process from which the data were obtained.
This randomness in the data must be taken into account when we interpret graphical and numerical summaries. Our conclusions should not be dependent on features that are specific to our particular data but would (probably) be different if the data were collected again.
Pages printed per ink jet cartridge
With a bigger data set, the dot plot, mean and standard deviation vary less between the different data sets.