r subset rows by criteria and by factor group

Question

I have this data.frame with a lot of NAs:

df <- data.frame(a = rep(letters[1:3], each = 3), 
                 b = c(NA, NA, NA, 1, NA, 3, NA, NA, 7))
df
> df
  a  b
1 a NA
2 a NA
3 a NA
4 b  1
5 b NA
6 b  3
7 c NA
8 c NA
9 c  7

I would like to subset this dataframe to obtain only factor group rows that have no less than two values, such as this:

  a  b
1 b 1
2 b NA
3 b 3

I have tried this function but it doesn't work:

subset(df, sum(!is.na(b)) < 1, by = a)

> [1] a b
<0 rows> (or 0-length row.names)

Any suggestion? (other packages solutions are welcome)

Answer 1

We can use data.table. Convert the 'data.frame' to 'data.table' (setDT(df)), grouped by 'a', if the sum of logical vector (i.e. non-NA elements - !is.na(b)) is greater than 1, then Subset the Data.table.

library(data.table)
setDT(df)[,if(sum(!is.na(b))>1) .SD , by = a]
#   a  b
#1: b  1
#2: b NA
#3: b  3

---

Or using dplyr, with the same logic, after grouping by 'a', we filter the rows.

library(dplyr)
df %>% 
    group_by(a) %>%
    filter(sum(!is.na(b))>1)
#      a     b
#  <fctr> <dbl>
#1      b     1
#2      b    NA
#3      b     3

---

Or in base R with ave

df[with(df, ave(b, a, FUN = function(x) sum(!is.na(x))>1)!=0),]

Answer 2

One way is using aggregate. Counting the number of elements which are not NA for a every unique a and then sub setting those rows from the data frame.

agg <- aggregate(b~a, df, function(x) length(!is.na(x)) >= 2)
df[df$a %in% agg[agg$b, "a"], ]

#  a  b
#4 b  1
#5 b NA
#6 b  3

---

Another option is using table.

df[df$a %in% names(which(table(df$a, is.na(df$b))[,1] > 1)), ]


#  a  b
#4 b  1
#5 b NA
#6 b  3