mixed models - Using custom made weights and structural weight in lme function in R -
i going use structural weight varindent(from = ~1 |sex)
, custom made vector of weights called w in lme
function in r. wondering if knows how combine weights in function. example (just made question)
n=25 data=data.frame( y = rnorm(n), sex = as.factor(sample(c('male','female','female'),size = n,replace = true)), age = as.factor(sample(c('1-1-2013','10-3-2013','1-5-2013'),size = n,replace = true)), area= as.factor(sample(c('a','b','a'),size = n,replace = true)) ) w = runif(n) my_weight_vector = w/sum(w) l = lme( fixed = y ~ sex + age, data = data, random = ~ 1 | area, weights = varident(form = ~1 | sex)+???my_weight_vector )
i believe want this:
data$v <- 1/my_weight_vector^2 l = lme( fixed = y ~ sex + age, data = data, random = ~ 1 | area, weights = varcomb(varident(form = ~1 | sex), varfixed(~ v)) )
varfixed
takes v
vector , uses fixed residual variance. demonstrating indeed results in desired weights:
varfun <- varfixed(~ v) varfun <- initialize(varfun, data) all.equal(varweights(varfun), my_weight_vector) #[1] true
Comments
Post a Comment