Daniel posted on Friday, September 17, 2004 - 7:16 am
I'm revising a paper in which I employed a quadratic growth curve in GGMM. Within one of my trajectories, geing female is associated positively with the linear trend and negatively with the quadratic trend. I know you suggest that interpreting quadratic trends is complicated, but given I have these results, what can I conclude? I also have a similar effect for baseline smoking in another trajectory, an ordered categorical covariate.
"Perhaps it helps to note that the linear trend has more to do with earlier development and the quadratic trend with later development."
This is interesting for me, because I am working on this issue right now. I am asking myself, if you can interpret an effect from the linear slope in one measure on the quadratic slope of another measure as an effect over time, if the factor loading for both processes are exactly the same.
Do you know any reference for this interpretation of the quadratic slope, I mean that it captures later time periods within one process?
I know of no references. But it seem reasonable if you think of the effect connected with the time scores. At t2 with timescore 1, the linear and the quadratic process have the same time score and therefore a 1 unit growth factor change has the same effect on the outcome. At t3, the time score is 2 for the linear and 4 for the quadratic, so the quadratic influence grows quicker.