proc traj

Gail

New Member
#1
Hello all, I use proc traj to create trajectories from repeated measures in a large sample.
Problem is the best model fit is when there is only 1 trajectory and fit becomes less good as the number of trajectories grows.
I would like to end up with 3 to 5 trajectories. Should I force this and ignore the model fit?
Thanks!
 
#2
Hello all, I use proc traj to create trajectories from repeated measures in a large sample.
Problem is the best model fit is when there is only 1 trajectory and fit becomes less good as the number of trajectories grows.
I would like to end up with 3 to 5 trajectories. Should I force this and ignore the model fit?
Thanks!
No. If you've specified the model and interpretted the output correctly, and there's no evidence of unobserved heterogenity in response, why would you think trying to force it would be a good idea?
 

Gail

New Member
#3
Thanks for your reply. I needed at least two trajectories because my aim was to examine the association between different trajectories and a dependent variable...