R. Problem with a loop through character strings - r

Question

I create this function to summarize the results of a glm:
outcome_forest<-function(mod,var,sd){
x<-summary(mod)
y<-x$coefficients
x_df<-as.data.frame(y)
x_df$Estimate<-x_df[var,1]/sd
x_df$ci_min<-x_df[var,1]-x_df[var,2]/sd
x_df$ci_max<-x_df[var,1]+x_df[var,2]/sd
return(x_df[var,c(1,5,6,4)])
}
Now I have different glm models:
mod_1<-glm(y_1~x_1+c_1+c_2,data=data_1, family = binomial)
mod_2<-glm(y_1~x_2+c_1+c_2,data=data_1, family = binomial)
I want to create a loop in order to pass my function to these two models:
thelist<-c("mod_1","mod_2")
sd<-c(0.58,0.98)
results<-list()
for(i in thelist){
for(j in sd){
results[[i]]<-outcome_forest(i,2,j)
}
}
I obtained the followin error
Error: $ operator is invalid for atomic vectors
I’m guessing this is happening because of the quota marks in model_1 and model_2. But these quotes are needed in order to create a thelist vector which just those names and no the results of both regression models.
How can I fix this issue?

Answer 1

You are correct, outcome_forest takes mod as an object, and you have it as a string c("mod_1","mod_2"). To make R evaluate that string as an object you need to use eval(parse(text=...)):
thelist<-c("mod_1","mod_2")
sd<-c(0.58,0.98)
results<-list()
for(i in thelist){
for(j in sd){
results[[i]]<-outcome_forest(eval(parse(text=i)),2,j)
}
}
But this approach isn't too god, its better to place the "mod"'s in a list and looping trough that:
l = list(mod_1, mod_2)
thelist<-c("mod_1","mod_2")
sd<-c(0.58,0.98)
results<-list()
for(i in 1:2){
for(j in sd){
results[[thelist[i]]]<-outcome_forest(l[[i]],2,j)
}
}
Or avoid creating thelist by naming results later:
sd<-c(0.58,0.98)
results<-list()
for(i in 1:2){
for(j in sd){
results[[i]]<-outcome_forest(l[[i]],2,j)
}
}
names(results) = c("mod_1","mod_2")