根据r中的NA值对实例进行分组

Question

I am reading a csv file and unfortunately my dataframe has many missing values. A small snip is as following:

df <- data.frame(Size= c(800, 850, 1100, 1200, 1000), 
                 Value= c(900, NA, 1300, 1100, NA),
                 Location= c(NA, 'midcity', 'uptown', NA, 'Lakeview'),
                 Num1 = c(2, NA, 3, 2, NA),
                 Num2 = c(2,3,3,1,2),
                 Rent= c('y', 'y', 'n', 'y', 'n'))

I want to predict some of the results using weka but I can't do it if I have multiple attributes missing. I know that I should be using the function is.na but I am not sure in what way it can be done because so far I used it only for summing and counting.

Edit: For an example, in this file I have missing values at 4 out of the 5 instances. Instances 2 and 5 share the same missing attributes (B and D), while instances 1 and 4 share the same missing value as well (C). What I'd like to get is a dataframe that consists out of those instances so I could export them into files and run analysis on those files individually. An example of an output could be

> A

> B

Edit 2:

I want to save the splits and so far I tried this:

write.csv(split(temp, index), file = "C:/Users/Nikita/Desktop/splits.csv", row.names=FALSE)

But it writes all the splits in one line. Is there a way to separate them by a line?

Edit 3:

My steps are:

data <- read.csv("location")
index <- apply(is.na(data)*1, 1,paste, collapse = "")
s <- split(data, index)
lapply(s, function(x) {names(x) <- names(data);x})
big.data <- do.call(rbind, s)
write.csv(big.data, file = "location", row.names=FALSE)

Am I missing something?

Answer 1

df[!is.na(df$Value), ]
  Size Value Location Num1 Num2 Rent
1  800   900     <NA>    2    2    y
3 1100  1300   uptown    3    3    n
4 1200  1100     <NA>    2    1    y

And

df[is.na(df$Value), ]
  Size Value Location Num1 Num2 Rent
2  850    NA  midcity   NA    3    y
5 1000    NA Lakeview   NA    2    n

In the future, please create a reproducible example so that users do not have to create a data frame by hand from your question. Pictures are not as helpful.

Data

df <- data.frame(Size= c(800, 850, 1100, 1200, 1000), 
                 Value= c(900, NA, 1300, 1100, NA),
                 Location= c(NA, 'midcity', 'uptown', NA, 'Lakeview'),
                 Num1 = c(2, NA, 3, 2, NA),
                 Num2 = c(2,3,3,1,2),
                 Rent= c('y', 'y', 'n', 'y', 'n'))

To combine it all use lapply since split creates a list:

lapply(split(temp, index), write.csv, file = "C:/Users/Nikita/Desktop/splits.csv", row.names=FALSE)

With a for loop:

s <- split(temp, index)
for (i in 1:length(s)) {
  write.csv(s[i], file = paste0("C:/Users/Nikita/Desktop/", i, "splits.csv"), row.names=FALSE)
}

Answer 2

Recreating your example data:

df <- data.frame(Size= c(800, 850, 1100, 1200, 1000), 
                 Value= c(900, NA, 1300, 1100, NA),
                 Location= c(NA, 'midcity', 'uptown', NA, 'Lakeview'),
                 Num1 = c(2, NA, 3, 2, NA),
                 Num2 = c(2,3,3,1,2),
                 Rent= c('y', 'y', 'n', 'y', 'n'))

Now, splitting your data according to the pattern of NA as you want:

# This generates an index with 1 for a column with NA and 0 otherwise
index <- apply(is.na(df)*1, 1,paste, collapse = "")

# This splits the data.frame according to the index
split(df, index)
$`000000`
  Size Value Location Num1 Num2 Rent
3 1100  1300   uptown    3    3    n

$`001000`
  Size Value Location Num1 Num2 Rent
1  800   900     <NA>    2    2    y
4 1200  1100     <NA>    2    1    y

$`010100`
  Size Value Location Num1 Num2 Rent
2  850    NA  midcity   NA    3    y
5 1000    NA Lakeview   NA    2    n

Notice that the first element "000000" comprises all the observations with complete cases. Then "001000" comprises all observations where column 3 (location) is missing. And so on.