Variable in a function is not used as the value

Question

I am using the package robust.arima in R, which works fine when I call it in a script. However, I want to organize my files and therefore call robust arima in a function. Here all of a sudden the variable is not found. Let me give an example

# Works fine
ts_list <- rnorm(100)
arima.rob(ts_list~1)

# Breaks down
get_rob_estimate <- function(x){
    out <- arima.rob(x~1)
    return(out)

ts_list <- rnorm(100)
get_rob_estimate(ts_list)

Error in eval(formula[[2]]) : object 'x' not found

Does anyone know what's going on? I think the problem looks similar to R : Pass argument to glm inside an R function , but I still can't seem to figure it out and I am curious how R processes these functions?

Edit

Okay for the basic option I understand it now, but I don't get why it works. What if I have

check_func <- function(ind_ts){
  out <- substitute(arima.rob(ind_ts~1))
  return(eval(out))
}

analyze_ts <- function(){
  df <- mvrnorm(100, mu=c(0,0,0), Sigma=diag(c(1,1,1)))
  p <- list()
  for (i in ncol(df)){
    sel <- df[,i]
    check_func(sel)
    p <- append(p, sel)
  }
  return(p)
}

analyze_ts()

I then get the error

Error in eval(formula[[2]]) : object 'sel' not found

How does it work? What is going on here? I just want my list to go as a list in my function, shouldn't be so hard right? Does not matter how many functions it goes through?

Answer 1

Using substitute()

get_rob_estimate <- function(x) {
  out <- substitute(robustarima::arima.rob(x ~ 1))
  return(eval(out))
}


get_rob_estimate(ts_list)
# Call:
# robustarima::arima.rob(formula = ts_list ~ 1)
# 
# Regression Coefficients:
#   (Intercept) 
#        0.1032 
# 
# Degrees of freedom: 100 total; 99 residual
# Innovations standard deviation: 0.9832 
# 
# Number of outliers detected:  1
# 
# Outlier index
# [1] 59
# 
# Outlier type
# [1] "AO"
# 
# Outlier impact
# [1] -3.0963
# 
# Outlier t-statistics
# [1] 3.1493

edit

You can write your Arima wrapper correctly like so:

analyze_ts <- function(){
  df <- MASS::mvrnorm(100, mu=c(0, 0, 0), Sigma=diag(c(1, 1, 1)))
  for (i in seq_len(ncol(df))) {
    sel <- df[,i]
    sel <- check_func(sel)
    p <- append(p, sel)
  }
  return(p)
}

Better using lapply

analyze_ts <- function() {
  df <- MASS::mvrnorm(100, mu=c(0, 0, 0), Sigma=diag(c(1,1,1)))
  return(lapply(seq_len(ncol(df)), \(i) check_func(df[,i])))
}

Usage:

set.seed(42) ## for sake of reproducibility
analyze_ts()

---

Data:

set.seed(42)
ts_list <- rnorm(100)