Given a model with two iid random effects, I'd like to take a linear combination of the standard deviation of the random effects, rather than levels of the random effects, in INLA.
Here an example, although the inla.make.lincomb is obviously incorrect.
library('INLA')
df <- lme4::sleepstudy
df$Days <- as.factor(df$Days)
df$Subject1 <- df$Subject
m1 <- inla(Reaction ~ f(Subject, model = "iid") +
f(Days, model = "iid"), data = df)
# This is incorrect, obviously, I'm looking for sd(Subject) + sd(Days)
lc <- inla.make.lincomb(Subject = c(1), Days = c(1))
mlc <- inla.hyperpar(inla(Reaction ~ f(Subject, model = "iid") +
f(Days, model = "iid"), data = df, lincomb = lc))
