An example of a simple hierarchical structure is people nested in families. This structure has two levels: people are at level one (the lower level), and they are grouped within families at level two (the higher level).

The reason for taking the family structure into account in our analysis is because we believe that individuals from the same family are likely to be more similar than people from different families. This may be because of shared social characteristics (e.g. household income), environmental characteristics (e.g. housing conditions) or genetic predisposition.

Other examples of two level structures are students in schools or patients in hospitals. The lower level units (students or patients) are nested within higher level units (schools or hospitals).

A feature worth noting is that multilevel modelling does not require the data to be balanced i.e. the number of people in each family does not have to be the same.

Other Hierarchical Data Structures
Often social data are only available at a level above the individual – for example household or area. Such data may still have a hierarchical structure.

For example, we may have information collected on households, and the households are nested within areas. In other words, households are our level one units and areas are at level two.

In many studies we are interested in the growth or development of individuals over time. For example, we may measure the height of the same children at different ages or the voting behaviour of adults over time.

These data again form two-level hierarchies with measurement occasions at level one nested within individuals at level two.

In the above example, person 2 may have been measured on five occasions; person 1 three times and person 3 just twice. Recall that multilevel modelling does not require a balanced structure, so can accommodate a varying number of occasions for each individual.

In many situations we will find that the data structure requires more than two levels. For example, we may have pupils within classes within schools, people within households within areas, or repeated measures on pupils within classes within schools.

Next Section: Research Questions