Error in xj[i] : only 0's may be mixed with negative subscripts

[Limin_Chen]

Hi Everyone,
I wrote an R function, when I call the function, it gave me the following error message:

wtMedian <- function(data, n){
chr <- unique(data[, which(colnames(data)=="chromosome")])
df <- data.frame(Col1 = character(),
                 Col2 = integer(),
                 Col3 = integer(),
                 Col4 = character(),
                 Col5 = double(),
                 Col6 = double(),
                 Col7 = double(), stringsAsFactors = FALSE)
weighted_m <- c() # new vector to store weighted median
for (ch in chr){
   data1 <- data %>% filter(chromosome %in% ch) # subset certain chromosome
  # nch <- length(data1$log2) # the $ will cause error when calling the function
  nch <- length(data1[,which(colnames(data)=="log2")])
  
  for (i in c(1:n)){
    #m <- median(data1[c(1:(n+i-1)),6])
    dat <- data1[c(1:(n+i-1)), which(colnames(data1)=="log2")] # data set for calculating weighed median
    w <- data1[c(1:(n+i-1)), which(colnames(data1)=="weight")] # weight set for calculating weighted median
    m <- weightedMedian(dat,w) # calculate weighted median
    weighted_m <- append(weighted_m, m)
  }
  for (i in c((n+1):(nch-n))){
    dat <- data1[c((i-n):(i+n-1)), which(colnames(data1)=="log2")]
    w <- data1[c((i-n):(i+n-1)), which(colnames(data1)=="weight")]
    m <- weightedMedian(dat,w)
    weighted_m <- append(weighted_m, m)
  }
  for (i in c((nch-n+1):nch)){
    dat <- data1[c((i-n):nch), which(colnames(data1)=="log2")]
    w <- data1[c((i-n):nch), which(colnames(data1)=="weight")]
    m <- weightedMedian(dat,w)
    weighted_m <- append(weighted_m, m)
  }
  df <- rbind(df, data1)
}
df$log2 <- weighted_m #assign weighted median to original dataframe
return(df)
}
df <- wtMedian(data,25)

Question2:
when I use $ in the function, there will be an error when calling the function, for example, if I replaced

chr <- unique(data[, which(colnames(data)=="chromosome")])

with

chr <- unique(data$chromosome). # the error message will be :

Error: [: $ operator is invalid for atomic vectors].
I used r which() to avoid that error, but causing the first error as mentioned.

However, if I don't run the above code in a function, everything will be fine.

Can anyone help me with the troubleshooting and explain what cause these errors, and why this specifically occurred in functions?

Greatly appreciated your help.

Best,
Lc

[AlexisW]

Question 1: hard to say for sure without knowing what is in data (please consider making a reprex), but it comes from using [ to select:

# let's create an example vector x
x <- 10:15

# you can use positive indexes to select one or several values
x[c(2,4)]
#> [1] 11 13

# you can use negative values to exclude these (and keep everything else)
x[ - c(2,4)]
#> [1] 10 12 14 15

# mixing the two is meaningless: do you want to keep or exclude the ones that are not mentioned?
x[c( -1, 3)]
#> Error in x[c(-1, 3)] : only 0's may be mixed with negative subscripts

_{more explanations here}

So in your function, somehow there is a combination of i and n where c((i-n):(i+n-1)) contains both positive and negative values, R can't guess what you want to do.

Question 2: looks like data is not a data.frame, according to the error message. It works for me:

wtMedian <- function(data){
  chr <- unique(data$chromosome)
  chr
}

my_data <- data.frame(chromosome = c("I","II"),
                       bla = 1:2)
wtMedian(my_data)
#> [1] "I"  "II"

^{Created on 2021-11-17 by the reprex package (v2.0.1)}

[Limin_Chen]

Hi AlexisW,

Thank for the explanation. That helps a lot in understanding the error. My issue was solved by changing those for loops into if statements. The error disappears.

Thanks a lot.

Best,
LC