[R-sig-ME] question about nested random effect

[Xuan Sun (student)]

Hi everyone

I do a research about the relationship between kid age and kids' height and other variables.
In my data, one mother has many kid, and a kid only have one mom.
I have consider the nested kid and mom. How could I consider the autocorrelation of each kid.
It is a 3 levels model, It is too hard for me.

Question 1: This is my modle now, how chould I correct it to consider the autocorrelation? In my opinion, I have to add a new level age which is nested in kid.id.(z-score here is standsrd children's height)
lmm_4<-lmer(weight.z.score ~ KID.AGE.YEARS
+ (1|MOM.ID/KID.ID),
data=data_calculate_without_na)

Question 2: When I add more variables, where should I but these variables? Is it right?
lmm_age_sib_WAZ<- gls(weight.z.score ~ KID.AGE.YEARS
+ as.factor(OLDER.SIBLINGS)
+ as.factor(YOUNG.SIBS)
+ KID.AGE.YEARS:OLDER.SIBLINGS
+ KID.AGE.YEARS:YOUNG.SIBS,
correlation = corCompSymm(form = ~ 1+KID.AGE.YEARS|MOM.ID/KID.ID),
data = HAZ_Sign_sib)

Qestion 3: when I want to fit it using gamm, what should I do to fit model consider these three levels. where should I add nested level in this model
gamm_age_sib_WAZ <-gamm(weight.z.score ~ s(KID.AGE.YEARS, bs = "ps",k=5)
+ s(as.factor(HAZ_Sign_sib$OLDER.SIBLINGS),
bs = "re",k=length(unique(HAZ_Sign_sib$OLDER.SIBLINGS)))
+ s(as.factor(HAZ_Sign_sib$YOUNG.SIBS),
bs = "re",k=length(unique(HAZ_Sign_sib$YOUNG.SIBS)))
+ te(KID.AGE.YEARS,by=(OLDER.SIBLINGS))
+ te(KID.AGE.YEARS,by=(YOUNG.SIBS)),
data = HAZ_Sign_sib)


If you can reply me, thank you so much!!


Best wishes

Xuan


	[[alternative HTML version deleted]]