I'm dealing with a problem which I had never thought about, so I am little uncomfortable choosing a solution by myself. I have data for cerebral arteries peak speed for a week after stroke. The problem is, whether blood speed at cerebral arteries at the same side of the stroke are higher than the speed at the arteries in the opposite side. I would like to make trajectories of these speeds for both homo and counter-lateral arteries of the stroke.
Looking at the data in the long format, there is something like this:
ID Stroke Day Speed Artery Side
32000000 L 1 37 R Counter
32000000 L 1 50 L Homo
32000000 L 2 46 R Counter
32000000 L 2 55 L Homo
32000000 L 3 55 R Counter
32000000 L 3 60 L Homo
At first, for each patient there are two measurements of speed, which is unususal for me. I am concerned because I figure there is an interdependence of the the speed within each patient, and there is a correlation of the speed over time. What would be a reasonable way to model these trajectories? If anyone is familiar with R, please give some package or function for me to look at.
You could simply treat this as a fixed effects OLS model, where you include a variable representing differences from T0 to T1. Check out Paul Allison's Sage book on fixed effects models.
Anil, the dependent variable is the speed, which I assume there is a correlation over time and an interdependence of the artery side.
William, Im not sure if I got right what you mean. This difference of time points would account for correlation? If so, the first time point of each patient would have value 0 of this difference? And the artery side would be inserted in the model as a fixed effect? Can you give a link to this book?
As far as I know, when you have dependent measurements, you use Hierarchical Linear Modeling.
This site might help - altough it is mainly about the MLwin Software for random effects , it also does contain more general material about multivariate models (with more than one response) and panel models
http://www.bristol.ac.uk/cmm/software/mlwin/mlwin-resources.html#multipleresp
http://www.bristol.ac.uk/cmm/software/mlwin/mlwin-resources.html#multipleresp
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis