statements not getting executed within a function, executing independently - r

Question

I have run into a strange issue (I am new to R). I have tried creating a function as follows:
library(ggplot2)
median_confidence_interval <- function(x) {
quart_list<-c()
return_data<-data.frame(lower_ci=0,median=0,upper_ci=0)
for(i in 1:1000){
y<-x[as.integer(runif(length(x), min = 1, max = length(x) + 1))]
median<-median(y)
quart_list=c(quart_list,median)
}
return_data$median<-median(quart_list)
return_data$lower_ci<-quantile(quart_list,probs=0.025)
return_data$upper_ci<-quantile(quart_list,probs=0.975)
p <- ggplot()
p <- p + geom_density(aes(x=x)) + geom_density(aes(x=quart_list))
p <- p + geom_vline(aes(xintercept = return_data$median, color='red'))
p <- p + geom_vline(aes(xintercept = return_data$lower_ci, color='blue'))
p <- p + geom_vline(aes(xintercept = return_data$upper_ci, color='green')) + coord_cartesian(xlim = c(min(x),max(x)))
png("density_confidence_internal.png")
plot(p)
dev.off()
return_data
}
In this code I am simply trying to create a plot and save it. Though I am able to execute each of these statements independently, outside the function, but not inside the function. The function compiles without errors but while running the function it says 'quart_list' not found.
If quart_list and return_data are present in the workspace, then I am able to execute the function and get the result. When I clear the workspace and execute the function, I run into the same error while running (not compiling).
Another issue is that when I call the function median_confidence_interval(x), it expects me to only provide 'x' as the argument, it doesn't take something like median_confidence_interval(possum$earconch). Why could that be?
Would someone please be able to point me in some direction?

Answer 1

The environment evaluation of ggplot objects is a little mysterious. However, if you remember that ggplot wants data.frames to be passed to the data argument and the values in aes should be columns of the data.frame you'll generally avoid issues.
To debug things like this, I find it helpful to insert print statements into the function to sort out how far through it I get. (see the commented lines)
Adjusting your function accordingly gives:
median_confidence_interval <- function(x) {
quart_list<-c()
return_data<-data.frame(lower_ci=0,median=0,upper_ci=0)
for(i in 1:1000){
y<-x[as.integer(runif(length(x), min = 1, max = length(x) + 1))]
median<-median(y)
# print ('inside for loop')
quart_list=c(quart_list,median)
}
# print('past for loop')
return_data$median<-median(quart_list)
return_data$lower_ci<-quantile(quart_list,probs=0.025)
return_data$upper_ci<-quantile(quart_list,probs=0.975)
# print('start of ggplot code')
foo=data.frame(q=quart_list, x=x)
p <- ggplot()
p <- p + geom_density(data=foo, aes(x=x)) + geom_density(data=foo, aes(x=q))
# print('past first quart_list reference in ggplot')
p <- p + geom_vline(data=return_data, aes(xintercept = median, color='red'))
p <- p + geom_vline(data=return_data, aes(xintercept = lower_ci, color='blue'))
p <- p + geom_vline(data=return_data, aes(xintercept = upper_ci, color='green')) + coord_cartesian(xlim = c(min(x), max(x)))
png("/tmp/density_confidence_internal.png")
plot(p)
dev.off()
return_data
}
Also, I think #DWin has a good point in his comment!