I am interested in changes in depression surrounding different life events and if personality traits (level-2 predictors) affect these changes in depression. In a given longitudinal data set, each subjects depression has been measured annually 9 times (or how often someone participated in the panel) as well as if a life event has occurred (measured monthly!). I created an event-centered time variable so that -1 represents the measurement occasion one month ahead of the event and +1 would be one month after the event.
I tried to start with an unconditional growth model with months surrounding the event as time predictor:
gcm.RS <- lmer(dep ~ 1 + event1.months + (1 + event1.months | id), data=event1)
Unfortunately I get the following warning message:
What does this mean? Has it something to do with my time variable?
In a next step I would like to test whether subjects differ in their change trajectories, and after that if personality explains the inter-individual differences in the intra-individual change curves. I thought to grand-mean center personality first. As this is my first-time use of lme4 I'm quite unsure if my procedure is right? Your comments on this are highly appreciated!
My ideas for these next models would be:
gcm.RI <- lmer(dep ~ 1 + event1.months + (1 | id), data=event1)
anova(gcm.RI, gcm.RS, refit=FALSE)
If subjects differ in their change curve I would stick to the RS model.
and for extraversion as an example:
attach(event1)
extrav.cgm <- extrav - mean(tapply(extrav, id, mean, na.rm=TRUE), na.rm=TRUE)
detach(event1)
gcm.CLI.extra <- lmer(dep ~ 1 + event1.months * extrav.cgm + (1 + event1.months | id), data=event1)
It would be an incredible help if you could let me know (1) what this warning messing means and subsequently what I am supposed to change and (2) if my other models are correct?
Many thanks in advance!
