在R中从另一个减去一个数字时出现NA

Question

I have data in the following format in R:

Col1 Col2 
1    1
2    1
4    0
0    0
2    2
.    .
.    .
.    .

I'm using the following script to work out the percentage differences between Col1 & Col2 in each row.

temp <- matrix(numeric(),dim(data)[1],1)

for (i in 1:dim(data)[1])
{
 temp[i,1]<- ((data[i,1]-data[i,2)/data[i,1])*100
}

For some reason my output file (temp) has some NA's in it. They are occurring even when 0-0. Does anyone know why it isn't just producing a 0 as opposed to NA? Some sums of 0-0 are producing a 0 as opposed to NA so I can't see any real pattern.

Any help would be much appreciated. Thanks,

Answer 1

You are making the classic division by zero error. R reports this as NaN - not a number, which is correct.

x <-structure(list(
  Col1 = c(1L, 2L, 4L, 0L, 2L), Col2 = c(1L, 1L, 0L, 0L, 2L)), 
  .Names = c("Col1", "Col2"), class = "data.frame", row.names = c(NA, -5L))

with(x, (Col2-Col1)/Col1)
[1]  0.0 -0.5 -1.0  NaN  0.0

---

One way to work around this is to use ifelse to return zero whenever Col==0 :

with(x, ifelse(Col1==0, 0, (Col2-Col1)/Col1))
[1]  0.0 -0.5 -1.0  0.0  0.0

---

If you don't want to use with , then write your code like this (more verbose but identical):

ifelse(x$Col1==0, 0, (x$Col2-x$Col1)/x$Col1)
[1]  0.0 -0.5 -1.0  0.0  0.0

Answer 2

For those entries, you're dividing by zero:

(0 - 0) / 0

Mathematically, the result of such a division is not defined . To indicate this, R evaluates 0/0 as NaN .

Here is a concise way to compute all percentags at once, returning zeroes for entries where both Col1 and Col2 are zero:

> data <- data.frame(Col1=c(1,2,4,0,2), Col2=c(1,1,0,0,2))
> pmax(100.0*(data$Col1-data$Col2)/data$Col1, 0, na.rm=T)
[1]   0  50 100   0   0