Grouped data
In this section, we analyse data sets that consist of a separate sample of numerical measurements from each of g groups. The data often arise from completely randomised experiments with g treatments.
Equivalently, the data may arise as bivariate data with a numerical response and a categorical 'explanatory' variable with g levels that identifies group membership.
Model for several groups
The values from each group are usually modelled as a random sample from some population. Normal distributions with the same standard deviations are often used as populations.
If the assumptions of a normal distribution and constant variance do not hold, a nonlinear transformation of the response may result in data for which the model is appropriate.
Illustration of the model