2 Data Structure for LMM

2.1 Data Examples

  • Patients data: Region, Health Board, Council, GP, Patients

  • School data: Area or Region, School, Class, Student

  • Repeated measures: GP, Patient, Multiple time points

See illustration below and note the structure of these different datasets


2.2 Data source

  • Complex experimental designs

  • Clinical trial

  • Repeated measures data

  • Longitudinal data

  • Clustered longitudinal data

  • Spatial data

  • Nested or Hierarchical data

  • Cross-classified data

These datasets are characterised by multiple strata or clusters where observations within a stratum are correlated


2.3 Data levels

  • Level 1:

    • The lowest level of data, usually the units of analysis in the study
    • The continuous response variable is always measured at Level 1
    • Example: Marks obtained by a student in the Math exam

  • Level 2:

    • The second level in the hierarchy
    • Level 2 observations represent the cluster of units
    • Example: Class representing units of students

  • Level 3:

    • The third level in the hierarchy
    • Level 3 observations represent the cluster of Level 2 units
    • Example: School representing units of classes

We measure continuous and categorical variables at different levels of the data and refer the variables as Level 1, 2 or 3 variables.



2.4 Illustration

Figure: Clustered data where individuals within a cluster are more likely to have similar measurements of the response variable



Figure: School data showing hierarchical or nested structure, also called multilevel structure



Figure: Repeated measures data where multiple measurements are recorded on each individual patient



Figure: Patient data showing hierarchical or nested structure at Area, GP and Patient level