I have repeated measures data (over time) for each of 40 individuals
with the number of observations per person varying greatly from 4 to
51. There are 578 measurements in total. There are no grouping
variables or other covariates.
I have been using the function carma in the library called growth to
fit a cubic polynomial in time with AR(1) serial correlation plus
measurement error. However, I have encountered a number of
problems:
1. Varying the initial estimates of the ARMA parameters results in
different fitted models and, in particular, the values of minus
log-likelihood. How can I choose a "best" model? Some fitted
models return NaN for some of the se's and/or correlations of the
parameter estiamtes so presumably these can be ruled out.
2. I am not sure how I specify that I want random effects in the
model and nor the nature of these (eg random intercept or random
intercept and slope, etc).
Any help would be greatly appreciated.
Peter
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !) To: r-help-request at
stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
