Longitudinal data - GEE

10 important questions on Longitudinal data - GEE

Which 5 correlation matrix structures can you choose from?

Exchangeable
Autoregressive
M-dependent
Unstructured
Independend

What is an exchangeable correlation matrix structure?

The same correlation coefficient for every time point

What is an Autoregressive correlation matrix structure?

You only calculate one regressioncoefficient (for 1 time interval).
2 time intervals = rho^2
3 time intervals = rho^3 
etc.
  • Higher grades + faster learning
  • Never study anything twice
  • 100% sure, 100% understanding
Discover Study Smart

What is an unstructured correlation matrix structure?

You calculate a correlation coefficient for each different time interval.

How much coefficients do you have to determine with an unstructured structure?

As much as there are obervations, but /2

What is a pro and a con for unstructured structure?

Pro: it is the best fitting the data
con: its very inefficient

What is the formula for approximating R2 with GEE analysis?

This
R2 = proportion of variance explained
variance (model) = scale parameter (on top of the output)
Variance (y) = Variance of your outcome. This is done by squaring your standard deviation of Y.

Which correlation structure is often used for GEE analysis with logistic regression? Why?

Exchangeable. This is because the outcome only has 2 options. It is way less likely to change over time for the same person. 1 correlationcoefficient is therefor good.

How can you decide the R2 of GEE with logistic regression?

You cant. There is no errors in logistic regression. Therefore you cannot calculate R2

Which GEE is the same as what kind of mixed model analysis?

A GEE with exchangeable structure is the same as mixed model with a random intercept

The question on the page originate from the summary of the following study material:

  • A unique study and practice tool
  • Never study anything twice again
  • Get the grades you hope for
  • 100% sure, 100% understanding
Remember faster, study better. Scientifically proven.
Trustpilot Logo