R: Aggregate error - ‘sum’ not meaningful for factors

user_12259903 0
Viewed 272

I've seen many similar questions on the website but somehow could not find an answer my problem. I have a data frame looking like this:

head(ftrade)
   Imports Value Exports Value  balance nacer2
1          7349        185712   178363     01
2       4772816      99763470 94990654     01
3       4772816      99763470 94990654     01
4       4772816      99763470 94990654     01
5       1022528       7880815  6858287     01
6       8295652        215331 -8080321

I want to aggregate my data by nacer2, while summing the values. My expected output would be like this:

    Imports Value Exports Value  balance nacer2
1         50000        100000    50000     01
2         50000        100000    50000     02
3         50000        100000    50000     03
4         50000        100000    50000     04
5         50000        100000    50000     05

where the values in the first three columns are the sum of the original data. I run the following:

ftrade <- do.call(data.frame, aggregate(cbind("Exports Value",
                                          "Imports Value",
                                           balance) ~ nacer2, 
                                           data = ftrade,
                                            sum))

which returns the error message : Error in Summary.factor(c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, : ‘sum’ not meaningful for factors

All the answers I've seen on this forum state that it is because one of the variables is a factor, and so summing does not make sense. I've checked, and none of my variables are factors:

str(ftrade)
'data.frame':   11963 obs. of  4 variables:
 $ Imports Value: num  7349 4772816 4772816 4772816 1022528 ...
 $ Exports Value: num  185712 99763470 99763470 99763470 7880815 ...
 $ balance      : num  178363 94990654 94990654 94990654 6858287 ...
 $ nacer2       : chr  "01" "01" "01" "01" ...

Since I am aggregating over nacer2, it should not be a problem that it is a character. I've try to convert everything in numeric values again, but nothing seems to solve my issue. I am not sure to understand what really is happening here. Am I missing something here ?

Thank you for your help, Clement

aggregate r
2 Answers

If you really need to use spaces in your variable names (you probably don't) then you need to use backticks to refer to them:

names(mtcars)[1] <- 'm p g'

aggregate(cbind(disp, 'm p g') ~ vs, mtcars, sum)

Error in Summary.factor(c(8L, 8L, 18L, 18L, 12L, 12L, 12L, 22L, 21L, 20L,  : 
  ‘sum’ not meaningful for factors

aggregate(cbind(disp, `m p g`) ~ vs, mtcars, sum)

  vs   disp m p g
1  0 5528.7 299.1
2  1 1854.4 343.8

Regardless, my advise is to not use spaces in your variable names.

user_43414400

I was able to do what I wanted using dplyr:

ftrade <- ftrade %>% 
  group_by(nacer2) %>%
  summarise(balance = sum(balance))

It does the work just fine, so I think we can consider the case closed. However, I am still curious to hear explanation about what exactly was happening here.

user_122599030
Related