The surveydata
package makes it easy to work with
typical survey data that originated in SPSS or other formats.
Specifically, the package makes it easy to include the question text as metadata with the data itself.
To track the questions of a survey, you have two options:
Neither of these options are ideal, since any subsetting of the survey data means you must keep track of the question metadata separately.
This package solves the problem by creating a new class,
surveydata
, and keeping the questions as an attribute of
this class. Whenever you do a subset operation, the metadata stays
intact.
In addition, the metadata knows if a question consists of a single column, or multiple columns. When creating a subset on the question name, the resulting object can be either a single column or multiple columns.
library(surveydata)
library(dplyr)
<- membersurvey %>% as.tbl() sv
## Warning: `as.tbl()` was deprecated in dplyr 1.0.0.
## ℹ Please use `tibble::as_tibble()` instead.
## Warning: 'as.tbl.surveydata' is deprecated.
## Use 'as_tibble' instead.
## See help("Deprecated")
sv
## # A tibble: 215 × 109
## id Q1_1 Q1_2 Q2 Q3_1 Q3_2 Q3_3 Q3_4 Q3_5 Q3_6 Q3_7 Q3_8 Q3_9
## <dbl> <dbl> <dbl> <ord> <fct> <fct> <fct> <fct> <fct> <fct> <fct> <fct> <fct>
## 1 3 8 2 2009 No No No No No No No No No
## 2 5 35 12 Befo… Yes No No No No No No No Yes
## 3 6 34 12 Befo… Yes Yes No No No Yes No No No
## 4 11 20 9 2010 No No No No No No No No No
## 5 13 20 3 2010 No No No No No No No No No
## 6 15 36 20 Befo… No Yes No No No No No No Yes
## 7 21 12 2.5 2009 Yes No No No No Yes Yes No No
## 8 22 11 0.5 2011 Yes Yes Yes Yes Yes No No No No
## 9 23 18 3 2008 Yes Yes Yes Yes Yes Yes No No Yes
## 10 25 24 8 2006 No No No Yes Yes Yes No No Yes
## # … with 205 more rows, and 96 more variables: Q3_10 <fct>, Q3_11 <fct>,
...
Notice from this summary that Question 2 has two columns,
i.e. Q2_1
and Q2_2
. You can extract both these
columns by simply referring to Q2
:
"Q2"] sv[,
## # A tibble: 215 × 1
## Q2
## <ord>
## 1 2009
## 2 Before 2002
## 3 Before 2002
## 4 2010
## 5 2010
## 6 Before 2002
## 7 2009
## 8 2011
## 9 2008
## 10 2006
## # … with 205 more rows
However, the subset of Q1
returns only a single
column:
"Q2"] sv[,
## # A tibble: 215 × 1
## Q2
## <ord>
## 1 2009
## 2 Before 2002
## 3 Before 2002
## 4 2010
## 5 2010
## 6 Before 2002
## 7 2009
## 8 2011
## 9 2008
## 10 2006
## # … with 205 more rows
Note that in both cases the surveydata
object doesn’t
return a vector - subsetting a surveydata
object always
returns a surveydata
object.
A surveydata object consists of:
A data frame with a row for each respondent and a column for each
question. Column names are typically names in the pattern
Q1
, Q2_1
, Q2_2
, Q3
-
where underscores separate the sub-questions when these originated in a
grid (array) of questions.
Question metadata gets stored in the `{variable.labels} attribute of the data frame. This typically contains the original questionnaire text for each question.
Information about the sub-question separator (typically an
underscore) is stored in the patterns
attribute.
Data processing a survey file can be tricky, since the standard
methods for dealing with data frames does not conserve the
variable.labels
attribute. The surveydata
package defines a surveydata
class and the following
methods that knows how to deal with the variable.labels
attribute:
as.surveydata
[.surveydata
[<-.surveydata
$.surveydata
$<-.surveydata
merge.surveydata
In addition, surveydata
defines the following convenient
methods for extracting and working with the variable labels:
varlabels
varlabels<-
First load the surveydata
package.
library(surveydata)
Next, create sample data. A data frame is the ideal data structure for survey data, and the convention is that data for each respondent is stored in the rows, while each column represents answers to a specific question.
<- data.frame(
sdat id = 1:4,
Q1 = c("Yes", "No", "Yes", "Yes"),
Q4_1 = c(1, 2, 1, 2),
Q4_2 = c(3, 4, 4, 3),
Q4_3 = c(5, 5, 6, 6),
Q10 = factor(c("Male", "Female", "Female", "Male")),
crossbreak = c("A", "A", "B", "B"),
weight = c(0.9, 1.1, 0.8, 1.2)
)
The survey metadata consists of the questionnaire text. For example, this can be represented by a character vector, with an element for each question.
To assign this metadata to the survey data, use the
varlabels()
function. This function assigns the
questionnaire text to the variable.labels
attribute of the
data frame.
varlabels(sdat) <- c(
"RespID",
"Question 1",
"Question 4: red", "Question 4: green", "Question 4: blue",
"Question 10",
"crossbreak",
"weight"
)
Finally, create the surveydata object. To do this, call the
as.surveydata()
function. The argument
renameVarlabels
controls whether the varlabels
get renamed with the same names as the data. This is an essential step,
and ensures that the question text remains in synch with the column
names.
<- as.surveydata(sdat, renameVarlabels = TRUE) sv
It is easy to extract specific questions with the [
operator. This works very similar to extraction of data frames. However,
there are two important differences:
surveydata
object, even if only a single column is
returned. This is different from the behaviour of data frames, where a
single column is simplified to a vector."Q1"] sv[,
## Q1
## 1 Yes
## 2 No
## 3 Yes
## 4 Yes
"Q4"] sv[,
## Q4_1 Q4_2 Q4_3
## 1 1 3 5
## 2 2 4 5
## 3 1 4 6
## 4 2 3 6
The extraction makes use of the underlying metadata, contained in the
varlabels
and pattern
attributes:
varlabels(sv)
## id Q1 Q4_1 Q4_2
## "RespID" "Question 1" "Question 4: red" "Question 4: green"
## Q4_3 Q10 crossbreak weight
## "Question 4: blue" "Question 10" "crossbreak" "weight"
pattern(sv)
## $sep
## [1] "_"
##
## $exclude
## [1] "other"
It is easy to query the surveydata object to find out which questions it contains, as well as which columns store the data for those questions.
questions(sv)
## [1] "id" "Q1" "Q4" "Q10" "crossbreak"
## [6] "weight"
which.q(sv, "Q1")
## [1] 2
which.q(sv, "Q4")
## [1] 3 4 5
The function question_text()
gives access to the
questionnaire text.
question_text(sv, "Q1")
## [1] "Question 1"
question_text(sv, "Q4")
## [1] "Question 4: red" "Question 4: green" "Question 4: blue"
Use question_text_common()
to retrieve the common text,
i.e. the question itself:
question_text_common(sv, "Q4")
## [1] "Question 4"
And use question_text_unique()
to retrieve the unique
part of the question, i.e. the sub-questions:
question_text_unique(sv, "Q4")
## [1] "red" "green" "blue"
surveydata
with dplyr
The surveydata
object knows how to deal with the
following dplyr
verbs:
select
filter
mutate
arrange
summarize
In every case the resulting object will also be of class
surveydata
.
The surveydata
object can make it much easier to work
with survey data.