如何使用R将两个或几个变量转换（计算）为一个？

Question

I'm having some difficulties merging two or several variables in my data. I'm able to do it in Excel but can't figure out how to perform the same thing in R.

Basically I want to create two combined variables using the variables below:

Data1: creating variable CombineA+B

country  year       A1         B1        **combineA1+B1**
USA      2002       0          0            0
USA      2003       1          1            2
USA      2004       NA         1            1
USA      2005       0          0            0
USA      2006       0          1            1
USA      2007       0          0            0
USA      2008       0          1            1
USA      2009       NA         NA           NA
USA      2010       0          1            1
USA      2011       NA         0            0
USA      2012       0          1            1
USA      2013       0          0            0
USA      2014       0          1            1

Creating the variable "combineA1+B1" seems simple, all I need to do is add those two (A1 and B1). In Excel this is very simple and I guess it is in R as well. However, NA values create problems when adding those two variables. So, how to create a combineA1+B1 variable like the one above?

If both A1 and B1 have NA's, then the combineA1+B1 should also have NA. If one has NA value and the other has 1 or 0 value, it should give the respective number (see for ex USA 2004).

I'd also like to create another combine variable: "combineA1+B1+C1+D1"

Data 2: creating variable "combineA1+B1+C1+D1"

country year    A1  B1  C1  D1  combineABCD
USA     2002    0   0   0   0   0
USA     2003    1   1   0   0   2
USA     2004    NA  1   0   0   1
USA     2005    0   0   0   0   0
USA     2006    0   1   0   0   1
USA     2007    0   0   0   0   0
USA     2008    0   1   1   0   2
USA     2009    NA  NA  NA  NA  NA
USA     2010    0   1   1   0   2
USA     2011    NA  0   0   0   0
USA     2012    0   1   1   0   2
USA     2013    0   0   0   0   0
USA     2014    0   1   1   0   2

I guess that once I know how to create the first combine variable I'll be able to do this as well. Although I'm not sure how all these NA's can be handled?

Grateful for all suggestions you can come up with to add these variable properly.

Answer 1

With a little bit of searching, I found this article . I take no credit for this code.

mysum <- function(x) if (all(is.na(x))) NA else sum(x, na.rm=T) 
df$combinedA1B1 <- apply(df[, c("A1", "B1")], 1, mysum)

df
#    country year A1 B1 combinedA1B1
# 1      USA 2002  0  0            0
# 2      USA 2003  1  1            2
# 3      USA 2004 NA  1            1
# 4      USA 2005  0  0            0
# 5      USA 2006  0  1            1
# 6      USA 2007  0  0            0
# 7      USA 2008  0  1            1
# 8      USA 2009 NA NA           NA
# 9      USA 2010  0  1            1
# 10     USA 2011 NA  0            0
# 11     USA 2012  0  1            1
# 12     USA 2013  0  0            0
# 13     USA 2014  0  1            1

Answer 2

To get R to drop NAs instead of propagating them through your calculation, many functions have an optional na.rm argument. It defaults to FALSE , but setting it to TRUE causes R to ignore NAs in your calculations:

> sum(1, NA)
[1] NA

> sum(1, NA, na.rm = TRUE)
[1] 1

However, passing this argument can cause tricky behavior when all of your arguments are NA , as R is still determined to ignore them:

> sum(NA, NA, na.rm = TRUE)
[1] 0

To get the kind of NA handling you want, you can define your own function:

my.sum <- function(...) {
    if(all(is.na(c(...)))) {
        return(NA)
    } else {
        return(sum(..., na.rm = TRUE))
    }
}

Once you've done that, you can zip your two columns together using mapply , like so:

data1$combine <- mapply(data1$A1, data1$B1, FUN = my.sum)

You may not have encountered ... yet for defining functions - its purpose is to take an arbitrary number of optional arguments and hold them to "pass on", in this case to c and sum .

Answer 3

Here is one with dplyr package:

df <- data.frame(country = rep("USA", 13),
             year = 2002:2014,
             A1 = c(0,1,NA,0,0,0,0,NA,0,NA,0,0,0),
             B1 = c(0,1,1,0,1,0,1,NA,1,0,1,0,1)
             n)

df <- df %>% mutate(combine = ifelse(is.na(A1), B1,
                           ifelse(is.na(B1), A1, A1 + B1)))