R: Error: new columns would leave holes after existing columns - r

Question

When running this code, I get the following error:
Error in `[<-.data.frame`(`*tmp*`, , i, value = list(x = 0.0654882985934691, :
new columns would leave holes after existing columns
I am trying to populate a data.frame with i number of columns, which with the output of the posted for loop should look like something like this (Excel example for convenience only):
The aim is to store the output of the loop in such a way that I can get the average of each column at a later stage.
What can be done to achieve this?
library(plyr)
library(forecast)
library(vars)
x <- rnorm(70)
y <- rnorm(70)
dx <- cbind(x,y)
dx <- as.ts(dx)
# Forecast Accuracy
j = 12 #Forecast horizon
k = nrow(dx)-j #length of minimum training set
prediction <- data.frame()
for (i in 1:j) {
trainingset <- window(dx, end = k+i-1)
testset <- window(dx, start = k+i, end = k+j)
fit <- VAR(trainingset, p = 2)
fcast <- forecast(fit, h = j-i+1)
fcastmean <- do.call('cbind', fcast[['mean']])
fcastmean <- as.data.frame(fcastmean)
prediction[,i] <- rbind(fcastmean[,1])
}
Edit
As per the comment below, I have edited the above code to specify the first variable of fcastmean.
The error I get has however changed as a result, now being:
Error in `[<-.data.frame`(`*tmp*`, , i, value = c(-0.316529962287372, :
replacement has 1 row, data has 0
Edit 2
Below is the minimum replicable version without any packages as requested in the comments. I believe that should be equivalent in terms of the question posed.
x <- rnorm(70)
y <- rnorm(70)
dx <- cbind(x,y)
dx <- as.ts(dx)
j = 12
k = nrow(dx)-j
prediction <- matrix(NA,j,j)
for (i in 1:j) {
fcast <- as.matrix(1:(j-i+1))
fcastmean <- fcast
prediction[,i] <- (fcastmean)
}

Answer 1

For your new example, try
sapply(1:j, function(i) `length<-`(1:(j-i+1), j))
The result is
[,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8] [,9] [,10] [,11] [,12]
[1,] 1 1 1 1 1 1 1 1 1 1 1 1
[2,] 2 2 2 2 2 2 2 2 2 2 2 NA
[3,] 3 3 3 3 3 3 3 3 3 3 NA NA
[4,] 4 4 4 4 4 4 4 4 4 NA NA NA
[5,] 5 5 5 5 5 5 5 5 NA NA NA NA
[6,] 6 6 6 6 6 6 6 NA NA NA NA NA
[7,] 7 7 7 7 7 7 NA NA NA NA NA NA
[8,] 8 8 8 8 8 NA NA NA NA NA NA NA
[9,] 9 9 9 9 NA NA NA NA NA NA NA NA
[10,] 10 10 10 NA NA NA NA NA NA NA NA NA
[11,] 11 11 NA NA NA NA NA NA NA NA NA NA
[12,] 12 NA NA NA NA NA NA NA NA NA NA NA
`length<-`(x, j) pads x with NA until it reaches a length of j.
You can replace 1:(j-i+1) with whatever function of i you want. In the OP's original example, I am guessing something like this will work (untested):
sapply(1:j, function(i){
trainingset <- window(dx, end = k+i-1)
# testset <- window(dx, start = k+i, end = k+j)
# ^ this isn't actually used...
fit <- VAR(trainingset, p = 2)
fcast <- forecast(fit, h = j-i+1)
`length<-`(fcast$mean, j)
})
function(i){...} is called an anonymous function and can be written like any other.