I am analyzing data on a study of the effects of Coronary Artery Bypass
Graft (CABG) on cognitive function, as measured by a score from an
objective test. I have 140 people who receive the CABG surgery and 92
controls, with four measurements of cognitive function over time (at 0,
3, 12 and 36 months). I have fitted a linear mixed model using lme with
a random intercept for subject and a random slope over time. I have
fixed effects for treatment, time and a learning effect, plus
interactions between time and treatment, and learning effect and
treatment. This model estimates one 2x2 covariance matrix for the
random effects.
My problem is that I wish to estimate a 2x2 random effects covariance
matrix for each treatment group. I have tried putting treatment and
time by treatment interaction terms into the random effects (assuming a
diagonal covariance matrix), but am unsure whether this is correct.
Could anyone recommend another approach?
Thanks for your help,
Sarah Barry, MS
Research Associate
Johns Hopkins Department of Biostatistics
Bloomberg School of Public Health
Phone: 410-614-1892
FAX: 410-955-0958
[[alternative HTML version deleted]]
