[R] Subscripting problem with is.na()

Bert Gunter bgunter.4567 at gmail.com
Fri Jun 24 17:37:05 CEST 2016


As Petr and Don have shown you, changing NA to 0 is unnecessary to get
what you want. However, recoding to 0 may be OK, as NA has a specific
meaning in this context, and you are just adding an extra code to a
factor for a different level.

But it still might cause you trouble later. One of R's strengths is
it's ability to simply deal with NA's -- most of the time anyway .For
example note that you would have to make sure these columns are
factors (*not numerics*), if you wanted to, say, investigate how
category of closing related to other covariates via e.g. multinomial
logistic regression or even just to tabulate the "closed" categories.
Keeping NA as NA allows R's built-in facilities to simply handle (e.g.
omit) the data for the "still open" cases, but you will have to do it
explicitly yourself if you code to 0. That seems to be asking for
trouble to me.

As always, contrary views welcome. This discussion still seems on
(r-help) topic to me, but if not, please say so.

Cheers,
Bert

Bert Gunter

"The trouble with having an open mind is that people keep coming along
and sticking things into it."
-- Opus (aka Berkeley Breathed in his "Bloom County" comic strip )


On Fri, Jun 24, 2016 at 12:14 AM,  <G.Maubach at gmx.de> wrote:
> Hi Bert,
>
> many thanks for all your help and your comments. I learn at lot this way.
>
> My question was about is.na() at the first sight but the actual task looks like this:
>
> I have two variables in my customer data that signal if the customer accout was closed by master data management or by sales. Say these variables are closed_mdm and closed_sls. They contain NA if the customer account is still open or a closing code from "01" to "08" if the customer account was closed and why.
>
> For my analysis I need a variable that combines the two variables closed_mdm and closed_sls to set a filter easily on those who are closed not matter what the reason was nor who closed the account.
>
> As I always encounter problems when dealing with ifelse statements and NA I decided to merge these two variables to one variable containing 0 = not closed and 1 = closed. In my context this seems to be - at least to me - a reasonable approach.
>
> Replacement of missing values and merging the variables is the easiest way for me.
>
> -- cut --
>
> cust_id <- c(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20)
> closed_mdm <- c("01", NA, NA, NA, "08", "07", NA, NA, "05", NA, NA, NA, "04", NA, NA, NA, NA, NA, NA, NA)
> closed_sls <- c(NA, "08", NA, NA, "08", "07", NA, NA, NA, NA, "03", NA, NA, NA, "05", NA, NA, NA, NA, NA)
>
> # 1st try
> ds_temp1 <- data.frame(cust_id, closed_mdm, closed_sls)
> ds_temp1
>
> ds_temp1$closed <- closed_mdm | closed_sls  # WRONG
>
> # 2nd try
> closed_mdm_fac1 <- as.factor(closed_mdm)
> closed_sls_fac1 <- as.factor(closed_sls)
>
> ds_temp2 <- data.frame(cust_id, closed_mdm_fac1, closed_sls_fac1)
> ds_temp2
>
> ds_temp2$closed <- ds_temp$closed_mdm_fac1 | ds_temp$closed_sls_fac1  # WRONG
>
> # 3rd try
> closed_mdm_num1 <- as.numeric(closed_mdm)  # OK
> closed_sls_num1 <- as.numeric(closed_sls)  # OK
>
> ds_temp3 <- data.frame(cust_id, closed_mdm_num1, closed_sls_num1)
> ds_temp3
>
> ds_temp3$closed <- ds_temp$closed_mdm_num1 | ds_temp$closed_sls_num1  # WRONG
>
> # 4th try
> ds_temp4 <- ds_temp3
> ds_temp4
>
> # Does not run due to not allowed NA in subscripts
> ds_temp4[is.na(ds_temp4$closed_mdm_num1), ds_temp4$closed_mdm_num1] <- 0
> ds_temp4[is.na(ds_temp4$closed_sls_num1), ds_temp4$closed_sls_num1] <- 0
>
> # 5th try
> ds_temp4$closed_mdm_num1 <- ifelse(is.na(ds_temp4$closed_mdm_num1), 1, 0)
> ds_temp4$closed_sls_num1 <- ifelse(is.na(ds_temp4$closed_sls_num1), 1, 0)
> ds_temp4
>
> ds_temp4$closed <- ifelse(ds_temp4$closed_mdm_num1 == 1 | ds_temp4$closed_sls_num1 == 1, 1, 0)
> ds_temp4
>
> -- cut --
>
> Is there a better way to do it?
>
> Kind regards
>
> Georg
>
>
>> Gesendet: Donnerstag, 23. Juni 2016 um 23:55 Uhr
>> Von: "Bert Gunter" <bgunter.4567 at gmail.com>
>> An: "David L Carlson" <dcarlson at tamu.edu>
>> Cc: "R Help" <r-help at r-project.org>
>> Betreff: Re: [R] Subscripting problem with is.na()
>>
>> ... actually, FWIW, I would say that this little discussion mostly
>> demonstrates why the OP's request is probably not a good idea in the
>> first place. Usually, NA's should be left as NA's to be dealt with
>> properly by R and packages. In biological measurements, for example,
>> NA's often mean "below the ability to reliably measure." Biologists
>> with whom I've worked over many years often want to convert these to 0
>> or omit the cases, both of which lead to biased estimates and/or
>> underestimates of variability and excess claims of "statistical
>> significance" (for those who belong to this religious persuasion). One
>> should never say never, but I suspect that there are relatively few
>> circumstances where the conversion the OP requested is actually wise.
>>
>> Feel free to ignore/reject such extraneous comments of course.
>>
>> Cheers,
>> Bert
>>
>>
>> Bert Gunter
>>
>> "The trouble with having an open mind is that people keep coming along
>> and sticking things into it."
>> -- Opus (aka Berkeley Breathed in his "Bloom County" comic strip )
>>
>>
>> On Thu, Jun 23, 2016 at 12:14 PM, David L Carlson <dcarlson at tamu.edu> wrote:
>> > Good point. I did not think about factors. Also your example raises another issue since column c is logical, but gets silently converted to numeric. This would seem to get the job done assuming the conversion is intended for numeric columns only:
>> >
>> >> test <- data.frame(a=c(1,NA,2), b = c("A","b",NA), c= rep(NA,3))
>> >> sapply(test, class)
>> >         a         b         c
>> > "numeric"  "factor" "logical"
>> >> num <- sapply(test, is.numeric)
>> >> test[, num][is.na(test[, num])] <- 0
>> >> test
>> >   a    b  c
>> > 1 1    A NA
>> > 2 0    b NA
>> > 3 2 <NA> NA
>> >
>> > David C
>> >
>> > -----Original Message-----
>> > From: Bert Gunter [mailto:bgunter.4567 at gmail.com]
>> > Sent: Thursday, June 23, 2016 1:48 PM
>> > To: David L Carlson
>> > Cc: Ivan Calandra; R Help
>> > Subject: Re: [R] Subscripting problem with is.na()
>> >
>> > Not in general, David:
>> >
>> > e.g.
>> >
>> >> test <- data.frame(a=c(1,NA,2), b = c("A","b",NA), c= rep(NA,3))
>> >
>> >> is.na(test)
>> >          a     b    c
>> > [1,] FALSE FALSE TRUE
>> > [2,]  TRUE FALSE TRUE
>> > [3,] FALSE  TRUE TRUE
>> >
>> >> test[is.na(test)]
>> > [1] NA NA NA NA NA
>> >
>> >> test[is.na(test)] <- 0
>> > Warning message:
>> > In `[<-.factor`(`*tmp*`, thisvar, value = 0) :
>> >   invalid factor level, NA generated
>> >
>> >> test
>> >   a    b c
>> > 1 1    A 0
>> > 2 0    b 0
>> > 3 2 <NA> 0
>> >
>> >
>> > The problem is the default conversion to factors and the replacement
>> > operation for factors. So:
>> >
>> >> test <- data.frame(a=c(1,NA,2), b = I(c("A","b",NA_character_)), c= rep(NA,3))
>> >> class(test$b)
>> > [1] "AsIs"  ## so NOT a factor
>> >
>> >> test[is.na(test)] <- 0 # now works as you describe
>> >> test
>> >   a b c
>> > 1 1 A 0
>> > 2 0 b 0
>> > 3 2 0 0
>> >
>> > Of course the OP (and you) probably had a data frame of all numerics
>> > in mind, so the problem doesn't arise. But I think one needs to make
>> > the distinction and issue clear.
>> >
>> > Cheers,
>> > Bert
>> >
>> >
>> >
>> >
>> >
>> > Bert Gunter
>> >
>> > "The trouble with having an open mind is that people keep coming along
>> > and sticking things into it."
>> > -- Opus (aka Berkeley Breathed in his "Bloom County" comic strip )
>> >
>> >
>> > On Thu, Jun 23, 2016 at 8:46 AM, David L Carlson <dcarlson at tamu.edu> wrote:
>> >> The function is.na() returns a matrix when applied to a data.frame so you can easily convert all the NAs to 0's:
>> >>
>> >>> ds_test
>> >>    var1 var2
>> >> 1     1    1
>> >> 2     2    2
>> >> 3     3    3
>> >> 4    NA   NA
>> >> 5     5    5
>> >> 6     6    6
>> >> 7     7    7
>> >> 8    NA   NA
>> >> 9     9    9
>> >> 10   10   10
>> >>> is.na(ds_test)
>> >>        var1  var2
>> >>  [1,] FALSE FALSE
>> >>  [2,] FALSE FALSE
>> >>  [3,] FALSE FALSE
>> >>  [4,]  TRUE  TRUE
>> >>  [5,] FALSE FALSE
>> >>  [6,] FALSE FALSE
>> >>  [7,] FALSE FALSE
>> >>  [8,]  TRUE  TRUE
>> >>  [9,] FALSE FALSE
>> >> [10,] FALSE FALSE
>> >>> ds_test[is.na(ds_test)] <- 0
>> >>> ds_test
>> >>    var1 var2
>> >> 1     1    1
>> >> 2     2    2
>> >> 3     3    3
>> >> 4     0    0
>> >> 5     5    5
>> >> 6     6    6
>> >> 7     7    7
>> >> 8     0    0
>> >> 9     9    9
>> >> 10   10   10
>> >>
>> >> -------------------------------------
>> >> David L Carlson
>> >> Department of Anthropology
>> >> Texas A&M University
>> >> College Station, TX 77840-4352
>> >>
>> >> -----Original Message-----
>> >> From: R-help [mailto:r-help-bounces at r-project.org] On Behalf Of Ivan Calandra
>> >> Sent: Thursday, June 23, 2016 10:14 AM
>> >> To: R Help
>> >> Subject: Re: [R] Subscripting problem with is.na()
>> >>
>> >> Thank you Bert for this clarification. It is indeed an important point.
>> >>
>> >> Ivan
>> >>
>> >> --
>> >> Ivan Calandra, PhD
>> >> Scientific Mediator
>> >> University of Reims Champagne-Ardenne
>> >> GEGENAA - EA 3795
>> >> CREA - 2 esplanade Roland Garros
>> >> 51100 Reims, France
>> >> +33(0)3 26 77 36 89
>> >> ivan.calandra at univ-reims.fr
>> >> --
>> >> https://www.researchgate.net/profile/Ivan_Calandra
>> >> https://publons.com/author/705639/
>> >>
>> >> Le 23/06/2016 à 17:06, Bert Gunter a écrit :
>> >>> Sorry, Ivan, your statement is incorrect:
>> >>>
>> >>> "When you use a single bracket on a list with only one argument in
>> >>> between, then R extracts "elements", i.e. columns in the case of a
>> >>> data.frame. This explains your errors. "
>> >>>
>> >>> e.g.
>> >>>
>> >>>> ex <- data.frame(a = 1:3, b = letters[1:3])
>> >>>> a <- 1:3
>> >>>> identical(ex[1], a)
>> >>> [1] FALSE
>> >>>
>> >>>> class(ex[1])
>> >>> [1] "data.frame"
>> >>>> class(a)
>> >>> [1] "integer"
>> >>>
>> >>> Compare:
>> >>>
>> >>>> identical(ex[[1]], a)
>> >>> [1] TRUE
>> >>>
>> >>> Why? Single bracket extraction on a list results in a list; double
>> >>> bracket extraction results in the element of the list ( a "column" in
>> >>> the case of a data frame, which is a specific kind of list). The
>> >>> relevant sections of ?Extract are:
>> >>>
>> >>> "Indexing by [ is similar to atomic vectors and selects a **list** of
>> >>> the specified element(s).
>> >>>
>> >>> Both [[ and $ select a **single element of the list**. "
>> >>>
>> >>>
>> >>> Hope this clarifies this often-confused issue.
>> >>>
>> >>>
>> >>> Cheers,
>> >>> Bert
>> >>> Bert Gunter
>> >>>
>> >>> "The trouble with having an open mind is that people keep coming along
>> >>> and sticking things into it."
>> >>> -- Opus (aka Berkeley Breathed in his "Bloom County" comic strip )
>> >>>
>> >>>
>> >>> On Thu, Jun 23, 2016 at 7:34 AM, Ivan Calandra
>> >>> <ivan.calandra at univ-reims.fr> wrote:
>> >>>> My statement "Using a single bracket '[' on a data.frame does the same as
>> >>>> for matrices: you need to specify rows and columns" was not correct.
>> >>>>
>> >>>>
>> >>>> When you use a single bracket on a list with only one argument in between,
>> >>>> then R extracts "elements", i.e. columns in the case of a data.frame. This
>> >>>> explains your errors.
>> >>>>
>> >>>> But it is possible to use a single bracket on a data.frame with 2 arguments
>> >>>> (rows, columns) separated by a comma, as with matrices. This is the solution
>> >>>> you received.
>> >>>>
>> >>>> Ivan
>> >>>>
>> >>>>
>> >>>> --
>> >>>> Ivan Calandra, PhD
>> >>>> Scientific Mediator
>> >>>> University of Reims Champagne-Ardenne
>> >>>> GEGENAA - EA 3795
>> >>>> CREA - 2 esplanade Roland Garros
>> >>>> 51100 Reims, France
>> >>>> +33(0)3 26 77 36 89
>> >>>> ivan.calandra at univ-reims.fr
>> >>>> --
>> >>>> https://www.researchgate.net/profile/Ivan_Calandra
>> >>>> https://publons.com/author/705639/
>> >>>>
>> >>>> Le 23/06/2016 à 16:27, Ivan Calandra a écrit :
>> >>>>> Dear Georg,
>> >>>>>
>> >>>>> You need to learn a bit more about the subsetting methods, depending on
>> >>>>> the object structure you're trying to subset.
>> >>>>>
>> >>>>> More specifically, when you run this: ds_test[is.na(ds_test$var1)]
>> >>>>> you get this error: "Error in `[.data.frame`(ds_test, is.na(ds_test$var1))
>> >>>>> : undefined columns selected"
>> >>>>>
>> >>>>> This means that R does not understand which column you're trying to
>> >>>>> select. But you're actually trying to select rows.
>> >>>>>
>> >>>>> Using a single bracket '[' on a data.frame does the same as for matrices:
>> >>>>> you need to specify rows and columns, like this:
>> >>>>> ds_test[is.na(ds_test$var1), ] ## notice the last comma
>> >>>>> ds_test[is.na(ds_test$var1), ] <- 0 ## works on all columns because you
>> >>>>> didn't specify any after the comma
>> >>>>>
>> >>>>> If you want it only for "var1", then you need to specify the column:
>> >>>>> ds_test[is.na(ds_test$var1), "var1"] <- 0
>> >>>>>
>> >>>>> It's the same problem with your 2nd and 4th tries (4th one has other
>> >>>>> problems). Your 3rd try does not change ds_test at all.
>> >>>>>
>> >>>>> HTH,
>> >>>>> Ivan
>> >>>>>
>> >>>>> --
>> >>>>> Ivan Calandra, PhD
>> >>>>> Scientific Mediator
>> >>>>> University of Reims Champagne-Ardenne
>> >>>>> GEGENAA - EA 3795
>> >>>>> CREA - 2 esplanade Roland Garros
>> >>>>> 51100 Reims, France
>> >>>>> +33(0)3 26 77 36 89
>> >>>>> ivan.calandra at univ-reims.fr
>> >>>>> --
>> >>>>> https://www.researchgate.net/profile/Ivan_Calandra
>> >>>>> https://publons.com/author/705639/
>> >>>>>
>> >>>>> Le 23/06/2016 à 15:57, G.Maubach at weinwolf.de a écrit :
>> >>>>>> Hi All,
>> >>>>>>
>> >>>>>> I would like to recode my NAs to 0. Using a single vector everything is
>> >>>>>> fine.
>> >>>>>>
>> >>>>>> But if I use a data.frame things go wrong:
>> >>>>>>
>> >>>>>> -- cut --
>> >>>>>>
>> >>>>>> var1 <- c(1:3, NA, 5:7, NA, 9:10)
>> >>>>>> var2 <- c(1:3, NA, 5:7, NA, 9:10)
>> >>>>>> ds_test <-
>> >>>>>>     data.frame(var1, var2)
>> >>>>>>
>> >>>>>> test <- var1
>> >>>>>> test[is.na(test)] <- 0
>> >>>>>> test  # NA recoded OK
>> >>>>>>
>> >>>>>> # First try
>> >>>>>> ds_test[is.na(ds_test$var1)] <- 0  # duplicate subscripts WRONG
>> >>>>>>
>> >>>>>> # Second try
>> >>>>>> ds_test[is.na("var1")] <- 0
>> >>>>>> ds_test$var1  # not recoded WRONG
>> >>>>>>
>> >>>>>> # Third try: to me the most intuitive approach
>> >>>>>> is.na(ds_test["var1"]) <- 0  # attempt to select less than one element in
>> >>>>>> integerOneIndex WRONG
>> >>>>>>
>> >>>>>> # Fourth try
>> >>>>>> ds_test[is.na(var1)] <- 0  # duplicate subscripts for columns WRONG
>> >>>>>>
>> >>>>>> -- cut --
>> >>>>>>    How can I do it correctly?
>> >>>>>>
>> >>>>>> Where could I have found something about it?
>> >>>>>>
>> >>>>>> Kind regards
>> >>>>>>
>> >>>>>> Georg
>> >>>>>>
>> >>>>>> ______________________________________________
>> >>>>>> R-help at 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.
>> >>>>>>
>> >>>>> ______________________________________________
>> >>>>> R-help at 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.
>> >>>>>
>> >>>> ______________________________________________
>> >>>> R-help at 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.
>> >>
>> >> ______________________________________________
>> >> R-help at 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.
>> >> ______________________________________________
>> >> R-help at 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.
>>
>> ______________________________________________
>> R-help at 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.



More information about the R-help mailing list