Kelvin Choi posted on Thursday, August 27, 2009 - 10:59 am
I have a longitudinal dataset with several age cohorts and modeled a growth curve for each one. I would like to obtain a predicted value at a particular follow-up time for all cohorts (e.g. at Round 4 of data collection), and compare these estimates to see if they are statistically different. How can I do it?
You should use Model constraint to define "New" parameters that express the estimated mean of the outcome at a certain time point in a certain group (cohort). The estimated mean is expressed in line with:
ytg = img + smg*xt;
where ytg is "New" parameter defined as the mean of y at time t for group g using the intercept growth factor img, slope growth factor smg, and time score xt. You define this for several groups and can then test if they are equal using Model Test.