Modelling two-group data
Numerical data from two groups are usually modelled as random samples from two underlying populations. Categorical data can be modelled in a similar way.
Typical data sets
We now show a few examples in which the difference between two sample proportions provides an estimate of the difference between the underlying probabilities.
Sample-to-sample variability
A simulation shows that the difference between the sample proportions is random and usually different from the difference between the population probabilities.
We will examine the distribution of p2 - p1 more carefully in the next page.