Apply filter to the table function

I'm looking for a way to execute a simple task faster than I am currently able to.

I want to use the table function in R on part of a dataframe. Of course it would be possible to first use subset and then table, but this is a bit tedious. (In my case, during a first inspection of the data, I want to check the frequency of NAs on individual variables in a multi-national survey for each of the 25 participating countries. So I'd need to create 25 subsets, make the table, and then remove the subsets again because I don't need them anymore.)

Here is some example data:

a <- c(1,1,1,1,1,2,2,2,2,2)
b <- c(1,3,99,99,2,3,2,99,1,1)
df <- cbind.data.frame(a,b)

And this is the workaround solution.

df1 <- subset(df, a == 1)
table(df1$b)
df2 <- subset(df, a == 2)
table(df2$b)
rm(df1, df2)

Is there a simpler way?

Also, I feel like I am spamming with ultra-basic questions like these. If anyone has a suggestion on how I could have found the answer directly I'd be happy to hear it. Other than trying some code myself, I googled terms like 'r apply filter to table', 'r filter table function', 'r table subset dataframe', etc.

Assuming 99 are your NAs then there is a way using purrr package, which I find is excellent to see how many NAs there are in each column:

library(purrr)
df |> 
  map_df(~sum(. == 99))

      a     b
  <int> <int>
1     0     3

Can you provide an example of the structure of the original data (multi-national survey)?

Probably you would be able to answer your question with a much tidier code using the package dplyr with functions such as

survey_data %>%
  select(column1, column2, country, etc) %>%  #choose your desired columns
  group_by(country) %>%
  summarise_all(funs(sum(is.na(.))))

You could split on your a variable and use lapply to use table on each list like this:

lapply(split(df, df$a), \(x) table(x))
#> $`1`
#>    b
#> a   1 2 3 99
#>   1 1 1 1  2
#> 
#> $`2`
#>    b
#> a   1 2 3 99
#>   2 2 1 1  1

^{Created on 2023-02-18 with reprex v2.0.2}

Just use it in an lapply.

alv <- unique(df$a)
lapply(alv, \(x) table(subset(df, a == x, b))) |> setNames(alv)
# $`1`
# b
# 1  2  3 99 
# 1  1  1  2 
# 
# $`2`
# b
# 1  2  3 99 
# 2  1  1  1 

However, it might be better to code 99 (and probably others) as NA,

df[] <- lapply(df, \(x) replace(x, x %in% c(99), NA))

and count the NAs in b for each individual a.

with(df, tapply(b, a, \(x) sum(is.na(x))))
# 1 2 
# 2 1 

Just use table() on the whole dataframe, and pull out the parts you want afterwards. You convert the a and b values to character values when indexing into the two-way table. For example,

a <- c(1,1,1,1,1,2,2,2,2,2)
b <- c(1,3,99,99,2,3,2,99,1,1)
df <- cbind.data.frame(a,b)
full <- table(df$a, df$b)
full["1",] # corresponds to subset a == 1
#>  1  2  3 99 
#>  1  1  1  2
full["2",] # corresponds to subset a == 2
#>  1  2  3 99 
#>  2  1  1  1
full[, "99"] # corresponds to subset b == 99
#> 1 2 
#> 2 1

^{Created on 2023-02-18 with reprex v2.0.2}