To make life a little more complex, because data is often insufficiently precise the manner we conceptualise them; therefore, there is a need to try and estimate the actual underlying data. And so estimation means new data is being created – that which may describe our reality but to an imprecise state. The estimations are based on actual empirical observations overlaid with some theory or stories about what it means for similar sort of situations and so on.
The more actual empirical observations, the statistical theory goes, the more we are able to refine our story (or the model) in order to improve the estimations. But the difficulty is that we can make mistakes in empirical observations; and certain assumptions have to be made about these mistakes and how much margin they end up constituting when we are dealing with large numbers of observations.
Incorporating intelligence about reality through theories or stories in different ways can help to deal with these mistakes as well. Being able to do so skillfully requires a strong understanding of both the statistical theories and also creativity in terms of introducing parameters into modelling.