I use the package Formula in R to create a conditional formula that can be updated.
(described here: How to update a conditional formula?)
By using the Formula package, my formula object gets two classes: "Formula" and "formula". Unfortunately, the "Formula" class does not work with the FENmlm package that I use for calculating my model. Is there a way to remove the "Formula" class and keep the "formula" class? (the "Formula" class is only needed for updating the models)
# Conditional formula
fml1 <- Formula::Formula(Petal.Width ~ Petal.Length | Species)
# Update conditional formula
fml2 <- update(fml, . ~ . + Sepal.Length)
# Class
class(fml2)
# [1] "Formula" "formula"
# Calculation
FENmlm::femlm(fml2, data = iris)
# Error in FENmlm::femlm(fml2, data = iris) :
# The argument 'fml' must be a formula.
# Try to delete "Formula" attribute
attributes(fml2)[[1]][[1]] <- NULL
# Error in attributes(fml2)[[1]][[1]] <- NULL : replacement has length zero
You can update the class of the formula with class <- -
fml1 <- Formula::Formula(Petal.Width ~ Petal.Length | Species)
# Update conditional formula
fml2 <- update(fml1, . ~ . + Sepal.Length)
class(fml2) <- 'formula'
FENmlm::femlm(fml2, data = iris)
#ML estimation, family = Poisson, Dep. Var.: Petal.Width
#Observations: 150
#Cluster sizes: Species: 3
#Standard-errors type: Standard
# Estimate Std. Error z value Pr(>|z|)
#Petal.Length 0.150326 0.259850 0.578512 0.562919
#Sepal.Length -0.012575 0.227379 -0.055302 0.955898
#---
#Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
# BIC: -147.53 Pseudo-R2: 0.19402
#Log-likelihood: 345.17 Squared Cor.: 0.94148