 # Trying to add values to an existing 1D table

There's probably an easy answer here I'm just missing, but I'm not finding it. Here's the scenario.

The specific goal I have is this: take a 1D table of frequencies [created with table()] that has many categories, and create a version that condenses the smaller categories into a combined "Other" category. So, for example, take the table

`````` A  B  C  D
10  6  2  1
``````

and convert it to

`````` A  B  Other
10  6      3
``````

by summing the entries C & D into the Other category.

More generally, as you'll see in the code sample, I'm trying to either concatenate two existing 1D tables together, or add a value/entry to an existing 1D table through a statement akin to `addtablevalue("Other" = 3)`. Ideally I'd find a way to use both approaches.

My code sample below includes what I thought would work but didn't, and an approach that get the job done but feel a little kludgy. Hoping there's a way that avoids a having to convert among types.

(Apologies if this is a bit long, because I took it straight from the notes I was keeping.)

``````# Create the full table of four entries and then a truncated version with
# the first two entries (categories A & B)
tmain <- table(c(rep("A", 10), rep("B", 6), "C", "C", "D"))
ttrunc <- tmain[1:2]

# Attempts to create the new table

# Approach 1 (A failure): Add a value to the truncated table by creating a
# new entry index, then add the name for that value
tnew <- ttrunc  # Copy our short table
tnew <- sum(tmain[3:4])   # Add a new value to the end of the table
names(tnew) <- "Other"   # Provide a name for that
# This initially appears to work - it's class is "table" and the right
# values are present, but str() shows a difference in structure details
# This has functional implications as barplot() fails
tnew   # Looks okay...
class(tnew)  # It is a table...
# But compare the structure now to the original structure of ttrunc
str(ttrunc)   # Gives 'table' int [1:2(1d)] 10 6 ...
str(tnew)   # Gives 'table' Named int [1:3] 10 6 3 ...
# Does applying as.table() fix it? No
tnew <- as.table(tnew)
str(tnew)  # Still the same non-standard table format
# A particular issue is that the altered format no longer works correctly in plots
barplot(ttrunc)  # The original table works
barplot(tnew)  # New one fails: 'height' must be a vector or a matrix

# Approach 2 (My work-around): Create a one-value table for Other then
# append it to ttrunc
# Create a new, one entry table of Other in the correct format
tother <- as.table(c("Other" = sum(tmain[3:4])))
str(tother)  # Format is a standard table

# How to append this to the table ttrunc?
# Not cbind() - It rotates ttrunc and then adds a column with two rows of
# the single value in tnew
cbind(ttrunc, tother)

# c() successfully combines the tables, but converts them into a vector
tnew <- c(ttrunc, tother)
str(tnew)
# (Using append() gives the same result)
# This does work with barplot()
barplot(tnew)
# And with prop.table(), although the output is still a vector
tprop <- prop.table(tnew)
str(tprop)

# To get true table format, this works, but feels like a kludge
# Can't we avoid having a vector in the first place?
tnew <- as.table(tnew)
``````

So as the notes suggest, it just seems there should be a way to do this that involves less steps and keeps the data in the normal `table` format throughout.

Thanks for any help!

``````tmain <- table(c(rep("A", 10), rep("B", 6), "C", "C", "D"))
ttrunc <- tmain[1:2]

ttrunc["Other"] <- sum(tmain[3:4])

ttrunc
#>     A     B Other
#>    10     6     3
``````

Thanks. Unfortunately, while that saves a step, it still has the same issue with producing a table in non-standard format:

``````tmain <- table(c(rep("A", 10), rep("B", 6), "C", "C", "D"))
ttrunc <- tmain[1:2]
tnew <- ttrunc  # Copy our short table
tnew["Other"] <- sum(tmain[3:4])   # Add a new value to the end of the table
# But compare the structure now to the original structure of ttrunc
str(ttrunc)   # Standard  'table' int [1:2(1d)] 10 6 ...
str(tnew)   # Still gives 'table' Named int [1:3] 10 6 3 ...
# Still does not work correctly in plots
barplot(tnew)  # Plot fails: 'height' must be a vector or a matrix
``````

Edit to add: It's frustrating, because it seems like it should work, but it doesn't! Maybe it's actually a bug?

See the FAQ: How to do a minimal reproducible example `reprex` for beginners. Having a `reprex` in a single chunk avoids potential tl;dr.

``````barplot(tmain)
``````

won't work either.

See the examples in `help("barplot")`

``````tmain <- table(c(rep("A", 10), rep("B", 6), "C", "C", "D"))
ttrunc <- tmain[1:2]

ttrunc["Other"] <- sum(tmain[3:4])

barplot(t(ttrunc[1:30]))
`````` Thanks again. The `t()` workaround does indeed work for the barplot, so thanks for that tip. I hadn't tried that approach.

But my real goal is to have the table end up in the standard format it has when first generated from a factor or character variable with `table()`. I actually found that `barplot(tmain)` works (see snippet below), and the combination of `table()` applied to a character/factor variable, with the result fed unaltered to `barplot`, is one I've used successfully in the past. But after adding the new entry that no longer works. I'm just trying to have the table stay in the exact same format when I add the additional value to to it. That is, have the original structure

'table' int [1:3(1d)] 10 6 3

• attr(*, "dimnames")=List of 1
..\$ : chr [1:3] "A" "B" "C"

rather than the modified

'table' Named int [1:3] 10 6 3

• attr(*, "names")= chr [1:3] "A" "B" "Other"

Notice that the modified table has lost the "(1d)" label from the vector, and the names are now stored directly as a character vector rather than being embedded in a list as in the original.

I should say, I don't expect you or anyone else to spend any real time trying to solve this when a few workarounds already exist. I was just hoping someone would know off the top of their head the "right" function that would do this in one step. The main motivation for that is the code will be used by students, and the fewer extra steps the less explanation is needed and the less confusing it is for them to follow.

``````tmain <- table(c(rep("A", 10), rep("B", 6), "C", "C", "D"))
barplot(tmain)   # This works for me
``````

As a guess, `table` was never intended to be subjected to subsetting. Although a subset returns a `table` object, the format doesn't completely carry over. A round trip through `table` cures that, but `as.table` doesn't.

``````tmain <- table(c(rep("A", 10), rep("B", 6), "C", "C", "D"))
ttrunc <- tmain[1:2]

# returns a table, but with different sub-obects
ttrunc["Other"] <- sum(tmain[3:4])
str(ttrunc)
#>  'table' Named int [1:3] 10 6 3
#>  - attr(*, "names")= chr [1:3] "A" "B" "Other"
rm(ttrunc)

# calling table again recreates the original aspects
ttrunc <- tmain[1:2]
ttrunc["Other"] <- sum(tmain[3:4])
ttrunc <- table(ttrunc)
str(ttrunc)
#>  'table' int [1:3(1d)] 1 1 1
#>  - attr(*, "dimnames")=List of 1
#>   ..\$ ttrunc: chr [1:3] "3" "6" "10"
rm(ttrunc)

# curiously, coercing to table doesn't
ttrunc <- tmain[1:2]
ttrunc["Other"] <- sum(tmain[3:4])
ttrunc <- as.table(ttrunc)
str(ttrunc)
#>  'table' Named int [1:3] 10 6 3
#>  - attr(*, "names")= chr [1:3] "A" "B" "Other"
``````

A go figure, I guess.

This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.

If you have a query related to it or one of the replies, start a new topic and refer back with a link.