5. Simulations

With a cross-table of a sample with two categorical variables, you can get an impression of the dependency of the variables. For this, you can use some measures like the relative risk and the chi square.

A sample of two non-correlated variables is independent, does not yet provide a perfect cross-table, and therefore not, for example, a relative risk of 1. It will be close, but that is a little too vague. You can create an image of that "neighborhood" using simulations by randomly taking many samples from the original original marginal data.

How this works and how to draw statistical conclusions is discussed in this chapter..