I've got a situation of longitudinal (per-second, actually) data collected differently by gender, so, e.g., the girls might systematically have shorter periods of observation than the boys, but all girls and all boys have constant measurement periods.
I want to estimate an LGM and be able to work with the trajectory parameters across gender.
What's the best way to do this? I'm imagining something like leaving all the girls' latter data missing and estimating a multiple group model, but I'm not sure what the pitfalls might be with the missing data estimation.
If I understand you correctly, both boys and girls were measured at the same time points, for example, Spring of each year. But they were observed for different periods of time at each time point. For example, the boys were observed for some behavior for ten seconds and girls were observed for some behavior for five seconds. Is this what you are saying?
I am still not clear on your situation. What comes to my mind is an observational setting where every five seconds for boys and every ten seconds for girls some type of behavior is noted. How may such notations are there for boys and how many for girls? Does each interval result in one variable? Please describe how the variables come about for boys and girls.
Ok, sorry about the confusion. Subjects come in for observation, and are hooked up for physiological measures. Boys do a task for, say, 80 seconds, and girls for 70 seconds. Data are recorded at each second.
So I want to be able to do growth modeling of the physio measures for both groups, and then test hypotheses relating the trajectory parameters across groups.
I guess you could try this with the last ten recordings missing for girls. Given that you have so many timepoints, you probably want to do this in long form rather than wide form using TYPE=TWOLEVEL and CLUSTER=ID;