Topic 4: Multilevel Model Flashcards
What is a multi-level trial
When observations are grouped into clusters
What is a multi-centre trial
Carried out at more than one medical insitution
What are the 3 types of multilevel design
Nested, crossed, repeated measure
What gets randomised in a nested design
The clusters are randomised. Therefore the clustering is imposed by the design
Why are multilevel trials different to other trials
Assumption of independence is violated, there is dependence among observations from that cluster. A larger sample size is needed for cluster randomisation, analysis must take clustering into account.
What is a crossed design termed
Background clustering
What is randomised in a nested design
Individuals, patients within clusters as opposed to the clusters themselves
What is a repeated measures design
Several measurements taken from the individual over a period of time
True/False: A nested and crossed design can have repeated measures
True
What is between cluster variation
Variation in the true means in outcome between clusters
What is within cluster correlation
Correlation between pairs of individuals in the same cluster - same for all pairs of individuals in the cluster
What is within cluster variance
How far each point is to the cluster specific mean. The variance within each cluster.
How is within cluster correlation measured
Intra-class correlation coefficient - rho
What is the intra-class correlation coefficient
The proportion of total variation in your outcome that is accounted for by clustering
What does an intra-class correlation coefficient of 1 mean
All individuals from the same cluster have the same outcome. Perfect correlation between individuals in the same cluster
