conditional mutate across multiple columns

[PossumHound]

Hi,

I would like to conditionally mutate numeric values across several columns; I have managed to do it and it works, but it looks super clumsy. Is it possible to write this in a nicer way?
Here an extract of the dataset:

I would like to conditionally change the value of all the numerical variables:

If the value is less then 10, I would like it replaced with NA (or ideally leave it blank if possible?)
If the value is more than 33, I would like it replaced with 33,
if the value is between 10and 33 then it should remain the same.

This is what I wrote:

tdata<- rawdata %>%
mutate(T1=ifelse(T1<10, NA, ifelse(T1<33, T1, 33)))%>%
mutate(T2=ifelse(T2<10, NA, ifelse(T2<33, T2, 33)))%>%
mutate(T3=ifelse(T3<10, NA, ifelse(T3<33, T3, 33)))%>%
mutate(C1=ifelse(C1<10, NA, ifelse(C1<33, C1, 33)))%>%
mutate(C2=ifelse(C2<10, NA, ifelse(C2<33, C2, 33)))%>%
mutate(C3=ifelse(C3<10, NA, ifelse(C3<33, C3, 33)))

Any idea how to simplify this?

Cheers

[technocrat]

Without a reprex, only a sketch will be offered

rawdata %>% mutate(T1 = case_when(
                         T1 < 10 ~ NA,
                         T1 >= 10 & T1 < 33 ~ T1,
                         T1 >= 30 ~ 33)
                  ),
           mutate ... [etc]

has greater legibility. Rather than repeating this code block for all six numeric variables, however, make it into a function and use mutate_if [numeric].

[PossumHound]

Thanks. How would I make it into function?

Cheers

[technocrat]

No assurance can be given that this will work in the absence of a reprex to test it, but the form of a function would be

tri_split <- function(x) {
  mutate(x = case_when(
    x < 10 ~ NA,
    x >= 10 & x < 33 ~ x,
    x >= 30 ~ 33)
  )
}

[PossumHound]

Thanks, I can't really figure out how to apply that...?

[technocrat]

And, without a reprex of your code, I can't either.

[PossumHound]

I have tried to reprex it and I can't get it to work. Also I can't really post more data than I already have. Not sure why you would need anything else.

Just to be clear, my code works fine:

tdata<- rawdata %>%
mutate(T1=ifelse(T1<10, NA, ifelse(T1<33, T1, 33)))%>%
mutate(T2=ifelse(T2<10, NA, ifelse(T2<33, T2, 33)))%>%
mutate(T3=ifelse(T3<10, NA, ifelse(T3<33, T3, 33)))%>%
mutate(C1=ifelse(C1<10, NA, ifelse(C1<33, C1, 33)))%>%
mutate(C2=ifelse(C2<10, NA, ifelse(C2<33, C2, 33)))%>%
mutate(C3=ifelse(C3<10, NA, ifelse(C3<33, C3, 33)))

It just seems like there should be a more efficient way to write it?

[andresrcs]

@PossumHound

Also I can't really post more data than I already have.

In order to give you a working solution, we need sample data in a copy/paste friendly format (we can't copy from a screenshot)

@PossumHound

Not sure why you would need anything else.

A proper reproducible example makes it much easier to understand your issue and figuring out how to help, also, it is a polite thing to do when asking questions on programming forums in general.

[PossumHound]

Someone helped me and wrote it this way:

calc1030 = function(x){
x = ifelse(x<10, NA, ifelse(x<33, x, 33))
return(x)
}

tdata<-rawdata %>% mutate_if(is.numeric, calc1030)

It looks much nicer now!

Another alternative would be

tdata<- rawdata %>% mutate(T1 = split_10_33(T1), T2 = split_10_33(T2), T3 = split_10_33(T3),
C1 = split_10_33(C1), C2 = split_10_33(C2), C3 = split_10_33(C3))

But again that's pretty convoluted.

Thank you anyway.

[PossumHound]

Sure no worries. I will try again next time. It keeps coming up with this error:

Rendering reprex...
pandoc.exe: \: openBinaryFile: does not exist (No such file or directory)
Error: pandoc document conversion failed with error 1

And I didn't really have time to deal with this on top of my actual problem.

Cheers for your help anyhow.

[andresrcs]

Using the reprex package makes your reproducible example look nicer, but it is not mandatory for providing one, it would be enough to provide sample data in a copy/paste friendly format (with dput() or datapasta::df_paste()) and the minimal runnable code that reproduces your issue.

[technocrat]

Not all of the data is needed. It doesn't even have to be real data. It just needs to illustrate the issue.

This has benefits:

1. It eliminates most sources of confusion.
2. A proposed solution can be illustrated.
3. It puts the burden where it belongs: on the person asking the question.
4. Questions that don't require reverse engineering attract more answers

[Dave_Armstrong]

The solution offered by PossumHound is great, but for completeness, the mutate_if() and in general the scoped versions of mutate() and summarise() have been superseded by the across() function. So, in this new regime, PossumHound's solution would be:

tdata<-rawdata %>% mutate(across(where(is.numeric), calc1030))

[ZykeZero]

Dave_Armstrong's answer is this is the correct and succinct way to do this.

The below is Dave's answer using the case_when() statement w/o assigning it to a function.

df %>% 
  summarise(
    across(where(is.numeric), 
           ~ case_when(
              .x < 10 ~ NA,
              .x >= 10 & .x < 33 ~ .x,
              .x >= 30 ~ 33)
    )
  )