I have a function I'm using in R that requires input to several parameters, once as a numeric (1) and as a character (NULL). The default is NULL.
I want to apply the function using all possible combinations of parameters, so I used expand.grid to try and create a dataframe which stores these. However, I am running into problems with creating an object that contains both numerics and characters in one column.
This is what I've tried:
comb<-expand.grid(c("NULL",1),c("NULL",1),stringsAsFactors=FALSE), which returns:
comb
Var1 Var2
1 NULL NULL
2 1 NULL
3 NULL 1
4 1 1
with all entries characters:
class(comb[1,1])
[1] "character"
If I now try and insert a numeric into a specific spot, I still receive a character:
comb[2,1]<-as.numeric(1)
class(comb[2,1])
[1] "character"
I've also tried it using stringsAsFactors=TRUE, or using expand.grid(c(0,1),c(0,1)) and then switching out the 0 for NULL but always have the exact same problem: whenever I do this, I do not get a numeric 1.
Manually creating an object using cbind and then inserting the NULL as a character also does not help. I'd be grateful for a pointer, or a work-around to running the function with all possible combinations of parameters.
As you have been told, generally speaking columns of data frames need to be a single type. It's hard to solve your specific problem, because it is likely that the solution is not really "putting multiple types into a single column" but rather re-organizing your other unseen code to work within this restriction.
As I suggested, it probably will be better to use the built in NA value as expand.grid(c(NA,1),c(NA,1)) and then modify your function to use NA as an input. Or, of course, you could just use some "special" numeric value, like -1, or -99 or something.
The related issue that I mentioned is that you really should avoid using the character string "NULL" to mean anything, since NULL is a special value in R, and confusion will ensue.
These sorts of strategies would all be preferable to mixing types, and using character strings of reserved words like NULL.
All that said, it technically is possible to get around this, but it is awkward, and not a good idea.
d <- data.frame(x = 1:5)
> d$y <- list("a",1,2,3,"b")
> d
x y
1 1 a
2 2 1
3 3 2
4 4 3
5 5 b
> str(d)
'data.frame': 5 obs. of 2 variables:
$ x: int 1 2 3 4 5
$ y:List of 5
..$ : chr "a"
..$ : num 1
..$ : num 2
..$ : num 3
..$ : chr "b"
Related
This seems pretty basic, but the number of verbs in the tidyverse is huge now and I don't know which package to look for this.
Here is the problem. I have a tibble
df <- tibble(f1 = factor(rep(letters[1:3],5)),
c1 = rnorm(15))
Now if I use the $ operator I can easily find out how many levels are in the factor.
nlevels(df$f1)
# [1] 3
But if I use the [] operator it returns an incorrect number of levels.
nlevels(df[,"f1"])
# [1] 0
Now if df is a data.frame and not a tibble the nlevels() function works with both the $ operator and the [] operator.
So does anyone know the tidyverse equivalent of nlevels() that works on both data.frames and tibbles?
Elaborating on the answer from timcdlucas (and the comments from r2evans), the issue here is the behavior of various forms of the extract operator, not the behavior of tibble. Why? a tibble is actually a kind of data.frame as illustrated when we use the str() function on a tibble.
> library(dplyr)
> aTibble <- tibble(f1 = factor(rep(letters[1:3],5)),
+ c1 = rnorm(15))
>
> # illustrate that aTibble is actually a type of data frame
> str(aTibble)
tibble [15 × 2] (S3: tbl_df/tbl/data.frame)
$ f1: Factor w/ 3 levels "a","b","c": 1 2 3 1 2 3 1 2 3 1 ...
$ c1: num [1:15] -0.5829 0.3682 1.1854 -0.6309 -0.0268 ...
There are four forms of the extract operator in R: [, [[, $, and #; as noted in What is the meaning of the dollar sign $ in R function?.
The first form, [ can be used to extract content form vectors, lists, matrices, or data frames. When used with a data frame (or tibble in the tidyverse), it returns an object of type data.frame or tibble unless the drop = TRUE argument is included, as noted in the question comments by r2evans.
Since the default setting of drop= in the [ function is FALSE, it follows that df[,"f1"] produces an unexpected or "wrong" result for the code posted with the original question.
library(dplyr)
aTibble <- tibble(f1 = factor(rep(letters[1:3],5)),
c1 = rnorm(15))
# produces unexpected answer
nlevels(aTibble[,"f1"])
> nlevels(aTibble[,"f1"])
[1] 0
The drop = argument is used when extracting from matrices or arrays (i.e. any object that has a dim attribute, as explained in help for the drop() function.
> dim(aTibble)
[1] 15 2
>
When we set drop = TRUE, the extract function returns an object of the lowest type available, that is all extents of length 1 are removed. In the case of the original question, drop = TRUE with the extract operator returns a factor, which is the right type of input for nlevels().
> nlevels(aTibble[,"f1",drop=TRUE])
[1] 3
The [[ and $ forms of the extract operator extract a single object, so they return objects of type factor, the required input to nlevels().
> str(aTibble$f1)
Factor w/ 3 levels "a","b","c": 1 2 3 1 2 3 1 2 3 1 ...
> nlevels(aTibble$f1)
[1] 3
>
> # produces expected answer
> str(aTibble[["f1"]])
Factor w/ 3 levels "a","b","c": 1 2 3 1 2 3 1 2 3 1 ...
> nlevels(aTibble[["f1"]])
[1] 3
>
The fourth form of the extract operator, # (known as the slot operator), is used with formally defined objects built with the S4 object system, and is not relevant for this question.
Conclusion: Base R is still relevant when using the Tidyverse
Per tidyverse.org, the tidyverse is a collection of R packages that share an underlying philosophy, grammar, and data structures. When one becomes familiar with the tidyverse family of packages, it's possible to do many things in R without understanding the fundamentals of how Base R works.
That said, when one incorporates Base R functions or functions from packages outside the tidyverse into tidyverse-style code, it's important to know key Base R concepts.
I think you might need to use [[ rather than [, e.g.,
> nlevels(df[["f1"]])
[1] 3
df[,"f1"] returns a tibble with one column. So you're doing nlevels on an entire tibble which doesn't make sense.
df %>% pull('f1') %>% nlevels
gives you what you want.
Preface:
I have seen this post:How to convert a factor to an integer\numeric without a loss of information? , but it does not really apply to the issue I am having. It addresses the issue of converting a vector in the form of factor to a numeric, but the issue I am having is larger than that.
Problem:
I am trying to convert a column in a dataframe from a factor to a numeric, while representing the dataframe using paste0. Here is an example:
aa=1:10
bb=rnorm(10)
dd=data.frame(aa,bb)
get(paste0("d","d"))[,2]=as.factor(get(paste0("d","d"))[,2])
(The actual code I am using requires me to use the paste0 function)
I get the error: target of assignment expands to non-language object
I am not sure how to do this, I think what is messing it up is the paste0 function.
First, this is not really a natural way to think about things or to code things in R. It can be done, but if you rephrase your question to give the bigger picture, someone can probably provide more natural ways of doing this in R. (Like the named lists #joran mentioned in the comment.)
With that said, to do this in R, you need to split apart the three steps you're trying to do in one line: get the data frame with the specified variable, make the desired column a factor, and then assign back to the variable name. Here I've wrapped this in a function, so the assignment needs to be made in pos=1 instead of the default, which would name it only within the function.
tof <- function(dfname, colnum) {
d <- get(dfname)
d[, colnum] <- factor(d[, colnum])
assign(dfname, d, pos=1)
}
dd <- data.frame(aa=1:10, bb=rnorm(10))
str(dd)
## 'data.frame': 10 obs. of 2 variables:
## $ aa: int 1 2 3 4 5 6 7 8 9 10
## $ bb: num -1.4824 0.7904 0.0258 1.2075 0.2455 ...
tof("dd", 2)
str(dd)
## 'data.frame': 10 obs. of 2 variables:
## $ aa: int 1 2 3 4 5 6 7 8 9 10
## $ bb: Factor w/ 10 levels "-1.48237228248052",..: 1 8 4 9 5 10 2 7 3 6
I am importing a 3 column CSV file. The final column is a series of entries which are either an integer, or a string in quotation marks.
Here are a series of example entries:
1,4,"m"
1,5,20
1,6,"Canada"
1,7,4
1,8,5
When I import this using read.csv, these are all just turned in to factors.
How can I set it up such that these are read as integers and strings?
Thank you!
This is not possible, since a given vector can only have a single mode (e.g. character, numeric, or logical).
However, you could split the vector into two separate vectors, one with numeric values and the second with character values:
vec <- c("m", 20, "Canada", 4, 5)
vnum <- as.numeric(vec)
vchar <- ifelse(is.na(vnum), vec, NA)
vnum
[1] NA 20 NA 4 5
vchar
[1] "m" NA "Canada" NA NA
EDIT Despite the OP's decision to accept this answer, #Andrie's answer is the preferred solution. My answer is meant only to inform about some odd features of data frames.
As others have pointed out, the short answer is that this isn't possible. data.frames are intended to contain columns of a single atomic type. #Andrie's suggestion is a good one, but just for kicks I thought I'd point out a way to shoehorn this type of data into a data.frame.
You can convert the offending column to a list (this code assumes you've set options(stringsAsFactors = FALSE)):
dat <- read.table(textConnection("1,4,'m'
1,5,20
1,6,'Canada'
1,7,4
1,8,5"),header = FALSE,sep = ",")
tmp <- as.list(as.numeric(dat$V3))
tmp[c(1,3)] <- dat$V3[c(1,3)]
dat$V3 <- tmp
str(dat)
'data.frame': 5 obs. of 3 variables:
$ V1: int 1 1 1 1 1
$ V2: int 4 5 6 7 8
$ V3:List of 5
..$ : chr "m"
..$ : num 20
..$ : chr "Canada"
..$ : num 4
..$ : num 5
Now, there are all sorts of reasons why this is a bad idea. For one, lots of code that you'd expect to play nicely with data.frames will not like this and either fail, or behave very strangely. But I thought I'd point it out as a curiosity.
No. A dataframe is a series of pasted together vectors (a list of vectors or matrices). Because each column is a vector it can not be classified as both integer and factor. It must be one or the other. You could split the vector apart into numeric and factor ( acolumn for each) but I don't believe this is what you want.
I am having rather lengthy problems concerning my data set and I believe that my trouble trace back to importing the data. I have looked at many other questions and answers as well as as many help sites as I can find, but I can't seem to make anything work. I am attemping to run some TTests on my data and have thus far been unable to do so. I believe the root cause is the data is imported as class NULL. I've tried to include as much information here as I can to show what I am working with and the types of issues I am having (in case the issue is in some other area)
An overview of my data and what i've been doing so far is this:
Example File data (as displayed in R after reading data from .csv file):
Part Q001 Q002 LA003 Q004 SA005 D106
1 5 3 text 99 text 3
2 3 text 2 text 2
3 2 4 3 text 5
4 99 5 text 2 2
5 4 2 1 text 3
So in my data, the "answers" are 1 through 5. 99 represents a question that was answered N/A. blanks represent unanswered questions. the 'text' questions are long and short answer/comments from a survey. All of them are stored in a large data set over over 150 Participants (Part) and over 300 questions (labled either Q, LA, SA, or D based on question with a 1-5 answer, long answer, short answer, or demographic (also numeric answers 0 thought 6 or so)).
When I import the data, I need to have it disregard any blank or 99 answers so they do not interfere with statistics. I also don't care about the comments, so I filter all of them out.
EDIT: data file looks like:
Part,Q001,Q002,LA003,Q004,SA005,D006
1,5,3,text,99,text,3
2,3,,text,2,text,2
etc...
I am using the following lines to read the data:
data.all <- read.table("data.csv", header=TRUE, sep=",", na.strings = c("","99"))
data <- data.all[, !(colnames(data.all) %in% c("LA003", "SA005")
now, when I type
class(data$Q001)
I get NULL
I need these to be Numeric. I can use summary(data) to get the means and such, but when I try to run ttests, I get errors including NULL.
I tried to turn this column into numerics by using
data<-sapply(data,as.numeric)
and I tried
data[,1]<-as.numeric(as.character(data[,1]))
(and with 2 instead of 1, but I don't really understand the sapply syntax, I saw it in several other answers and was trying to make it work)
when I then type
class(data$Q001)
I get "Error: $ operator is invalid for atomic vectors
If I do not try to use sapply, and I try to run a ttest, I've created subsets such as
data.2<-subset(data, D106 == "2")
data.3<-subset(data, D106 == "3")
and I use
t.test(data.2$Q001~data.3$Q001, na.rm=TRUE)
and I get "invalid type (NULL) for variable 'data.2$Q001'
I tried using the different syntax, trying to see if I can get anything to work, and
t.test(data.2$Q001, data.3$Q001, na.rm=TRUE)
gives "In is.na(d) : is.na() applied to non-(list or vector) of type 'NULL'" and "In mean.default(x) : argument is not numeric or logical: returning NA"
So, now that I think I've been clear about what I'm trying to do and some of the things I've tried...
How can I import my data so that numbers (specifically any number in a column with a header starting with Q) are accurately read as numbers and do not get a NULL class applied to them? What do I need to do in order to get my data properly imported to run TTests on it? I've used TTests on plenty of data in the past, but it has always been data I recorded manually in excel (and thus had only one column of numbers with no blanks or NAs) and I've never had an issue, and I just do not understand what it is about this data set that I can't get it to work. Any assistance in the right direction is much appreciated!
This works for me:
> z <- read.table(textConnection("Part,Q001,Q002,LA003,Q004,SA005,D006
+ 1,5,3,text,99,text,3
+ 2,3,,text,2,text,2
+ "),header=TRUE,sep=",",na.strings=c("","99"))
> str(z)
'data.frame': 2 obs. of 7 variables:
$ Part : int 1 2
$ Q001 : int 5 3
$ Q002 : int 3 NA
$ LA003: Factor w/ 1 level "text": 1 1
$ Q004 : int NA 2
$ SA005: Factor w/ 1 level "text": 1 1
$ D006 : int 3 2
> z2 <- z[,!(colnames(z) %in% c("LA003","SA005"))]
> str(z2)
'data.frame': 2 obs. of 5 variables:
$ Part: int 1 2
$ Q001: int 5 3
$ Q002: int 3 NA
$ Q004: int NA 2
$ D006: int 3 2
> z2$Q001
[1] 5 3
> class(z2$Q001)
[1] "integer"
The only I can think of is that your second command (which was missing some terminating parentheses and brackets) didn't work at all, you missed seeing the error message, and you are referring to some previously defined data object that doesn't have the same columns defined. For example, class(z$QQQ) is NULL following the above example.
edit: it appears that the original problem was some weird/garbage characters in the header that messed up the name of the first column. Manually renaming the column (names(data)[1] <- "Q001") seems to have fixed the problem.
Many surveys have codes for different kinds of missingness. For instance, a codebook might indicate:
0-99 Data
-1 Question not asked
-5 Do not know
-7 Refused to respond
-9 Module not asked
Stata has a beautiful facility for handling these multiple kinds of missingness, in that it allows you to assign a generic . to missing data, but more specific kinds of missingness (.a, .b, .c, ..., .z) are allowed as well. All the commands which look at missingness report answers for all the missing entries however specified, but you can sort out the various kinds of missingness later on as well. This is particularly helpful when you believe that refusal to respond has different implications for the imputation strategy than does question not asked.
I have never run across such a facility in R, but I would really like to have this capability. Are there any ways of marking several different types of NA? I could imagine creating more data (either a vector of length nrow(my.data.frame) containing the types of missingness, or a more compact index of which rows had what types of missingness), but that seems pretty unwieldy.
I know what you look for, and that is not implemented in R. I have no knowledge of a package where that is implemented, but it's not too difficult to code it yourself.
A workable way is to add a dataframe to the attributes, containing the codes. To prevent doubling the whole dataframe and save space, I'd add the indices in that dataframe instead of reconstructing a complete dataframe.
eg :
NACode <- function(x,code){
Df <- sapply(x,function(i){
i[i %in% code] <- NA
i
})
id <- which(is.na(Df))
rowid <- id %% nrow(x)
colid <- id %/% nrow(x) + 1
NAdf <- data.frame(
id,rowid,colid,
value = as.matrix(x)[id]
)
Df <- as.data.frame(Df)
attr(Df,"NAcode") <- NAdf
Df
}
This allows to do :
> Df <- data.frame(A = 1:10,B=c(1:5,-1,-2,-3,9,10) )
> code <- list("Missing"=-1,"Not Answered"=-2,"Don't know"=-3)
> DfwithNA <- NACode(Df,code)
> str(DfwithNA)
'data.frame': 10 obs. of 2 variables:
$ A: num 1 2 3 4 5 6 7 8 9 10
$ B: num 1 2 3 4 5 NA NA NA 9 10
- attr(*, "NAcode")='data.frame': 3 obs. of 4 variables:
..$ id : int 16 17 18
..$ rowid: int 6 7 8
..$ colid: num 2 2 2
..$ value: num -1 -2 -3
The function can also be adjusted to add an extra attribute that gives you the label for the different values, see also this question. You could backtransform by :
ChangeNAToCode <- function(x,code){
NAval <- attr(x,"NAcode")
for(i in which(NAval$value %in% code))
x[NAval$rowid[i],NAval$colid[i]] <- NAval$value[i]
x
}
> Dfback <- ChangeNAToCode(DfwithNA,c(-2,-3))
> str(Dfback)
'data.frame': 10 obs. of 2 variables:
$ A: num 1 2 3 4 5 6 7 8 9 10
$ B: num 1 2 3 4 5 NA -2 -3 9 10
- attr(*, "NAcode")='data.frame': 3 obs. of 4 variables:
..$ id : int 16 17 18
..$ rowid: int 6 7 8
..$ colid: num 2 2 2
..$ value: num -1 -2 -3
This allows to change only the codes you want, if that ever is necessary. The function can be adapted to return all codes when no argument is given. Similar functions can be constructed to extract data based on the code, I guess you can figure that one out yourself.
But in one line : using attributes and indices might be a nice way of doing it.
The most obvious way seems to use two vectors:
Vector 1: a data vector, where all missing values are represented using NA. For example, c(2, 50, NA, NA)
Vector 2: a vector of factors, indicating the type of data. For example, factor(c(1, 1, -1, -7)) where factor 1 indicates the a correctly answered question.
Having this structure would give you a create deal of flexibility, since all the standard na.rm arguments still work with your data vector, but you can use more complex concepts with the factor vector.
Update following questions from #gsk3
Data storage will dramatically increase: The data storage will double. However, if doubling the size causes real problem it may be worth thinking about other strategies.
Programs don't automatically deal with it. That's a strange comment. Some functions by default handle NAs in a sensible way. However, you want to treat the NAs differently so that implies that you will have to do something bespoke. If you want to just analyse data where the NA's are "Question not asked", then just use a data frame subset.
now you have to manipulate two vectors together every time you want to conceptually manipulate a variable I suppose I envisaged a data frame of the two vectors. I would subset the data frame based on the second vector.
There's no standard implementation, so my solution might differ from someone else's. True. However, if an off the shelf package doesn't meet your needs, then (almost) by definition you want to do something different.
I should state that I have never analysed survey data (although I have analysed large biological data sets). My answers above appear quite defensive, but that's not my intention. I think your question is a good one, and I'm interested in other responses.
This is more than just a "technical" issue. You should have a thorough statistical background in missing value analysis and imputation. One solution requires playing with R and ggobi. You can assign extremely negative values to several types of NA (put NAs into margin), and do some diagnostics "manually". You should bare in mind that there are three types of NA:
MCAR - missing completely at random, where P(missing|observed,unobserved) = P(missing)
MAR - missing at random, where P(missing|observed,unobserved) = P(missing|observed)
MNAR - missing not at random (or non-ignorable), where P(missing|observed,unobserved) cannot be quantified in any way.
IMHO this question is more suitable for CrossValidated.
But here's a link from SO that you may find useful:
Handling missing/incomplete data in R--is there function to mask but not remove NAs?
You can dispense with NA entirely and just use the coded values. You can then also roll them up to a global missing value. I often prefer to code without NA since NA can cause problems in coding and I like to be able to control exactly what is going into the analysis. If have also used the string "NA" to represent NA which often makes things easier.
-Ralph Winters
I usually use them as values, as Ralph already suggested, since the type of missing value seems to be data, but on one or two occasions where I mainly wanted it for documentation I have used an attribute on the value, e.g.
> a <- NA
> attr(a, 'na.type') <- -1
> print(a)
[1] NA
attr(,"na.type")
[1] -1
That way my analysis is clean but I still keep the documentation. But as I said: usually I keep the values.
Allan.
I´d like to add to the "statistical background component" here. Statistical analysis with missing data is a very good read on this.