I am trying to extract LMS constants from a gamlss object. I have the following gamlss object, where the best fitting family is the BCT.
m1tc <- gamlss(tc~pb(age), data = girls.only.lipid, family=BCT)
From the help page I got the following:
calculate transformed age
t.age <- fitted(lm(girls.only.lipid$age~fitted(m1tc))) t.age <- t.age - min(t.age) data.t <- data.frame(girls.only.lipid,t.age=t.age)
calculate final solution
girlstc1 <- gamlss(tc~pb(t.age), data=data.t,family=BCT()) library(AGD) girlstc2 <- extractLMS(fit = girlstc1, data=data.t, sex="F", grid="0-21y", dec = c(2,2,5))
This computes the LMS constants, however, they do not fit. When using them to calculate SDS values, these are way off (to large) when looking at my plots. If I use the NO family in the "girlstc1" function, it seems to fit much better. But would it be making the wrong assumptions, as the best fit for the function (using AIC) is still the BCT family? Or is something else wrong with the coding?