|
|
 |
| Random logv affected by a time varyin... |
 
|
| Message/Author |
|
|
| Jon Heron posted on Wednesday, September 25, 2019 - 8:45 am
|
|
|
Hi everyone
had a really interesting discussion with a colleague earlier and i'm wondering if its possible to shoe-horn the problem we discussed into DSEM, perhaps cross-classified style.
We have access to a number of intensive datasets linking sleep to aspects of mental health. For this particular question we might use a population who have daily ratings sleep and current mood over the course of at least a year.
The question is whether poor (poorer than average) sleep is followed by a period of increased mood lability, i.e. an increase in innovation variance (logv) for that participant. So logv would vary between people but also be affected by something that changes within people.
This feels like a bit of a no-go as I am trying to link sleep residuals to the variance of mood residuals. I'm wondering if I need to put this on hold for future Mplus releases so I can think of good/poor sleep as a switching between latent classes.
Sorry for the bombardment of ideas, we're preparing a bid for future work and scoping out what might be possible.
best wishes, Jon |
|
|
|
|
| Cross-classified analysis would be helpful to a point. If you look at slide 59 (and application slide 75) of my Parts 3 and 4 handout from Hopkins, you see that equation (3) has a between-subject alpha_i and a between-time alpha_t where alpha_t could be seen as a deviation at time point t (up or down) from the subject's mean level. So alpha_t could be used as a predictor on the Between time level. But current Mplus does not allow logv to vary over time in cross-classified so at most you could use alpha_t for sleep as a predictor of alpha_t for mood. |
|
|
| Jon Heron posted on Wednesday, September 25, 2019 - 11:56 pm
|
|
|
| Thanks Bengt that sounds useful |
|
|
| Back to top |
|
|
