Problems with the lme function with nested variables (any(notIntX <- ! apply(X, 2, const))

Question

I’m trying to adjust a mixed effects model in which I have covariables that are nested (Varx5 | Varx6) and are considered fixed effects.

However, I am trying to adjust the data and the following message appears:

library(nlme)
library(lme4)

dados$VarCat=as.factor(dados$VarCat)
dados$VarX5=as.factor(dados$VarX5)
dados$VarX6=as.factor(dados$VarX6)

model <- lme(log(Resp)~log(VarX1)+log(VarX2)+(VarX3)+(VarX4)+VarX5|VarX6 ,random = ~1|VarCat, 
                 dados, method="REML")

Error in if (any(notIntX <- !apply(X, 2, const))) { : 
  valor ausente onde TRUE/FALSE necessário

Answer 1

You are not specifying categorical variables in the appropriate way. In nlme:lme they stay out of the formula, in the option random. In the lme4:lmer they go directly in the formula, but bounded by parentheses.

dados <- read.csv2("https://docs.google.com/uc?id=1hKfYpcgAV4gdDPHXv_4EQpiO0Y31_ZVo&export=download")

model <- nlme::lme(
  log(Resp) ~ log(VarX1) + log(VarX2) + VarX3 + VarX4,
  random = ~ VarCat | VarX5:VarX6,
  data = dados)

model <- lme4::lmer(
  log(Resp) ~ log(VarX1) + log(VarX2) + VarX3 + VarX4 + (VarCat | VarX5:VarX6),
  data = dados)

The operator : is an example, I’m assuming you want the conditional effect. For cross-purpose use /. Check the details for formula in aid of lmer for different specifications.