[R] Question concerning side effects of treating invalid factor levels

[Sarah Goslee]

Hi Tibor,

No, you are misunderstanding the source of the problem. It has nothing
to do with factors.

Instead, it has to do with the inability of a vector to hold more than
one class.

You are using rbind() to add a new row to your data frame, but that
vector is being coerced to character. That's what is forcing your
numeric column to become character: you're adding a character to it.

> c("in", "V>N", round(runif(1, 7000, 16000), 0))
[1] "in"    "V>N"   "15709"

It has nothing whatsoever to do with factors or factor levels, and
would occur if you were adding it to a data frame with character
values.

If you want to mix types, you cannot use a vector.

c2 <- data.frame(P = "in", ANSWER = "V>N", RT = round(runif(1, 7000, 16000), 0))
> str(rbind(df, c2))
'data.frame': 7 obs. of  3 variables:
 $ P     : Factor w/ 4 levels "mit","mittels",..: 2 1 2 3 1 1 4
 $ ANSWER: Factor w/ 3 levels "OBJ>PP","PP>OBJ",..: 2 2 2 2 1 1 3
 $ RT    : num  10867 14808 11600 15881 8984 ...


Sarah

On Tue, Sep 20, 2022 at 8:45 AM Tibor Kiss via R-help
<r-help using r-project.org> wrote:
>
> Hi,
>
> this is a misunderstanding of my question. I wasn’t worried about invalid factor levels that produce NA. My question was why a column changes its class, which I thought was a side effect. If you add a vector containing one character string, the class of the whole vector becomes _chr_. And after this element has been added to a column, we have two NAs for the column which are factors, and a character string, which is responsible for the change of a numerical vector into a character string vector (see ?c, where you find: "The output type is determined from the highest type of the components in the hierarchy NULL < raw < logical < integer < double < complex < character < list < expression.“).
>
>
> Best
>
>
> Tibor
>
>
>
> > Am 19.09.2022 um 13:59 schrieb Ebert,Timothy Aaron <tebert using ufl.edu>:
> >
> > In your example code, the variable remains a class factor, and all entries are valid. The variables will behave as expected given the factor levels in the original dataframe.
> >
> > (At least on my system R 4.2, in RStudio, in Windows) R returns a couple of error messages warning me that I was bad.
> > What you get is NA for "not available", or "not appropriate" or a missing value. You gave the system an invalid factor level so it was entered as missing. If you get data that has a new factor level, you need to tell R to expect a new factor level first.
> >
> > levels(f1) <- c(levels(f1),"New Level")
> > levels(f1) <- c(levels(f1),c("NL1","NL2"))
> >
> >
> > Tim
> > -----Original Message-----
> > From: R-help <r-help-bounces using r-project.org> On Behalf Of Tibor Kiss via R-help
> > Sent: Monday, September 19, 2022 6:11 AM
> > To: r-help using r-project.org
> > Subject: [R] Question concerning side effects of treating invalid factor levels
> >
> > [External Email]
> >
> > Dear List members,
> >
> > I have tried now for several times to find out about a side effect of treating invalid factor levels, but did not find an answer. Various answers on stackexchange etc. produce the stuff that irritates me without even mentioning it.
> > So I am asking the list (apologies if this has been treated in the past).
> >
> > If you add an invalid factor level to a column in a data frame, this has the side effect of turning a numerical column into a column with character strings. Here is a simple example:
> >
> >> df <- data.frame(
> >        P = factor(c("mittels", "mit", "mittels", "ueber", "mit", "mit")),
> >        ANSWER = factor(c(rep("PP>OBJ", 4), rep("OBJ>PP", 2))),
> >        RT = round(runif(6, 7000, 16000), 0))
> >
> >> str(df)
> > 'data.frame':   6 obs. of  3 variables:
> > $ P     : Factor w/ 3 levels "mit","mittels",..: 2 1 2 3 1 1
> > $ ANSWER: Factor w/ 2 levels "OBJ>PP","PP>OBJ": 2 2 2 2 1 1
> > $ RT    : num  11157 13719 14388 14527 14686 ..
> >
> >> df <- rbind(df, c("in", "V>N", round(runif(1, 7000, 16000), 0)))
> >
> >> str(df)
> > 'data.frame':   7 obs. of  3 variables:
> > $ P     : Factor w/ 3 levels "mit","mittels",..: 2 1 2 3 1 1 NA
> > $ ANSWER: Factor w/ 2 levels "OBJ>PP","PP>OBJ": 2 2 2 2 1 1 NA
> > $ RT    : chr  "11478" "15819" "8305" "8852" ...
> >
> > You see that RT has changed from _num_ to _chr_ as a side effect of adding the invalid factor level as NA. I would appreciate understanding what the purpose of the type coercion is.
> >
> > Thanks in advance
> >
> >
> > Tibor
> > ______________________________________________
> > R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
> > https://nam10.safelinks.protection.outlook.com/?url=https%3A%2F%2Fstat.ethz.ch%2Fmailman%2Flistinfo%2Fr-help&data=05%7C01%7Ctebert%40ufl.edu%7C6ee1a1f50c14442beef508da9a301bde%7C0d4da0f84a314d76ace60a62331e1b84%7C0%7C0%7C637991828670135028%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=sNDYEJKhjSu%2FtrTIwZx5yVemKgDheQYXLrcQqJ2mOgo%3D&reserved=0
> > PLEASE do read the posting guide https://nam10.safelinks.protection.outlook.com/?url=http%3A%2F%2Fwww.r-project.org%2Fposting-guide.html&data=05%7C01%7Ctebert%40ufl.edu%7C6ee1a1f50c14442beef508da9a301bde%7C0d4da0f84a314d76ace60a62331e1b84%7C0%7C0%7C637991828670135028%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=AP%2B4fa5pvbGr3IfwdiQvjXwkOdY90CIWIWWWmpIHH7w%3D&reserved=0
> > and provide commented, minimal, self-contained, reproducible code.
>
>
>         [[alternative HTML version deleted]]
>
> ______________________________________________
> R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.



-- 
Sarah Goslee (she/her)
http://www.numberwright.com