library(surveydata)
# Create surveydata object
sdat <- data.frame(
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)
)
varlabels(sdat) <- c(
"RespID",
"Question 1",
"Question 4: red", "Question 4: green", "Question 4: blue",
"Question 10",
"crossbreak",
"weight"
)
sv <- as.surveydata(sdat, renameVarlabels = TRUE)
# Extract specific questions
sv[, "Q1"]
sv[, "Q4"]
# Query attributes
varlabels(sv)
pattern(sv)
# Find unique questions
questions(sv)
which.q(sv, "Q1")
which.q(sv, "Q4")
# Find question text
question_text(sv, "Q1")
question_text(sv, "Q4")
question_text_common(sv, "Q4")
question_text_unique(sv, "Q4")
# Basic operations on a surveydata object, illustrated with the example dataset membersurvey
class(membersurvey)
questions(membersurvey)
which.q(membersurvey, "Q1")
which.q(membersurvey, "Q3")
which.q(membersurvey, c("Q1", "Q3"))
question_text(membersurvey, "Q3")
question_text_unique(membersurvey, "Q3")
question_text_common(membersurvey, "Q3")
# Extracting columns from a surveydata object
head(membersurvey[, "Q1"])
head(membersurvey["Q1"])
head(membersurvey[, "Q3"])
head(membersurvey[, c("Q1", "Q3")])
# Note that the result is always a surveydata object, even if only one column is extracted
head(membersurvey[, "id"])
str(membersurvey[, "id"])
Run the code above in your browser using DataLab