Longitudinal Study

[mansi birla]

Hello all,
In a longitudinal study, if we assume linear model y = u + e , where e
is residual vector, assuming that e is normally distributed with mean
0 and variance covariance matrix V. Also, y is a vector of repeated
measurements on a subject. I have doubt whether the matrix V will be a
diagonal matrix or not , i.e., whether the error term will be
correlated or not? Please suggest.

[Peter Flom]

If you have repeated measurements on Y then the errors/residuals will not be
independent, which is another assumption of the linear model. One solution
is multilevel models (aka mixed models, hierarchical linear models, etc).
Another is generalized estimating equations.

Peter Flom
Peter Flom Consulting
http://www.statisticalanalysisconsulting.com/
http://www.IAmLearningDisabled.com
http://www.linkedin.com/in/peterflom

--
To post a new thread to MedStats, send email to MedS...@googlegroups.com .
MedStats' home page is http://groups.google.com/group/MedStats .
Rules: http://groups.google.com/group/MedStats/web/medstats-rules