🚀 Wrangle Workflow
1. Reshape Your Data
Outcome naming & order (important)
- List
.outcomesin the order questions were asked.
- If you have a repeated task, its outcome must be the last element.
- For base tasks (all but last), the function reads the digits in each name as the task id (e.g.,
"choice4","Q4","task04"→ task 4).
- The repeated base task is inferred from the first base outcome’s digits. The repeated outcome itself need not contain digits—only its position (last) matters.
- Specify the two exported response values with
.choice_map. Its names are the response strings stored in the outcome columns, and its values are the corresponding Qualtrics profile positions (1or2). For example, use.choice_map = c("Community A" = 1, "Community B" = 2)when those are the exported choices and the instrument confirms that Community A is profile 1. projoint cannot infer this mapping from the CSV.- Invalid labels, trailing whitespace, and missing choices now stop with an informative error. Retain missing choices only after review by setting
.allow_missing_choices = TRUE.
Specify response labels and profile positions
First inspect the response values that actually appear in your outcome columns. For example:
## [1] "Community A" "Community B"Then verify the relationship between those response values and the two profile positions using the Qualtrics instrument or QSF file. Write the verified relationship as:
.choice_map = c(
"exact response value for profile 1" = 1,
"exact response value for profile 2" = 2
)The strings on the left are not new profile names created by
reshape_projoint(). They must match the ends of the values
stored in every outcome column, including capitalization and whitespace.
The numbers on the right refer to the profile positions encoded in
columns such as K-1-1-* and K-1-2-*. Although
shorter suffixes such as "A" and "B" are
supported, using the complete exported response strings is clearer and
more auditable.
Example (Flipped Repeated Task)
outcomes <- paste0("choice", 1:8)
outcomes1 <- c(outcomes, "choice1_repeated_flipped")
out1 <- reshape_projoint(
.dataframe = exampleData1,
.outcomes = outcomes1,
.choice_map = c("Community A" = 1, "Community B" = 2),
.alphabet = "K",
.idvar = "ResponseId",
.repeated = TRUE,
.flipped = TRUE
)Key Arguments:
.outcomes: Outcome columns (include repeated task last).choice_map: Verified mapping from exact exported response values to Qualtrics profile positions 1 and 2.idvar: Respondent ID variable.alphabet: Variable prefix (“K”).repeated,.flipped: If repeated task exists and is flipped
2. Variations: Repeated vs. Non-Repeated
Not-Flipped Repeated Task
outcomes <- paste0("choice", 1:8)
outcomes2 <- c(outcomes, "choice1_repeated_notflipped")
out2 <- reshape_projoint(
.dataframe = exampleData2,
.outcomes = outcomes2,
.repeated = TRUE,
.flipped = FALSE
)No Repeated Task
3. The .fill Argument: Should You Use It?
Use .fill = TRUE to “fill” missing values based on IRR
agreement.
fill_FALSE <- reshape_projoint(
.dataframe = exampleData1,
.outcomes = outcomes1,
.fill = FALSE
)
fill_TRUE <- reshape_projoint(
.dataframe = exampleData1,
.outcomes = outcomes1,
.fill = TRUE
)Compare:
selected_vars <- c("id", "task", "profile", "selected", "selected_repeated", "agree")
fill_FALSE$data[selected_vars]## # A tibble: 6,400 Ă— 6
## id task profile selected selected_repeated agree
## <chr> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 R_00zYHdY1te1Qlrz 1 1 1 1 1
## 2 R_00zYHdY1te1Qlrz 1 2 0 0 1
## 3 R_00zYHdY1te1Qlrz 2 1 1 NA NA
## 4 R_00zYHdY1te1Qlrz 2 2 0 NA NA
## 5 R_00zYHdY1te1Qlrz 3 1 1 NA NA
## 6 R_00zYHdY1te1Qlrz 3 2 0 NA NA
## 7 R_00zYHdY1te1Qlrz 4 1 0 NA NA
## 8 R_00zYHdY1te1Qlrz 4 2 1 NA NA
## 9 R_00zYHdY1te1Qlrz 5 1 1 NA NA
## 10 R_00zYHdY1te1Qlrz 5 2 0 NA NA
## # â„ą 6,390 more rows## # A tibble: 6,400 Ă— 6
## id task profile selected selected_repeated agree
## <chr> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 R_00zYHdY1te1Qlrz 1 1 1 1 1
## 2 R_00zYHdY1te1Qlrz 1 2 0 0 1
## 3 R_00zYHdY1te1Qlrz 2 1 1 NA 1
## 4 R_00zYHdY1te1Qlrz 2 2 0 NA 1
## 5 R_00zYHdY1te1Qlrz 3 1 1 NA 1
## 6 R_00zYHdY1te1Qlrz 3 2 0 NA 1
## 7 R_00zYHdY1te1Qlrz 4 1 0 NA 1
## 8 R_00zYHdY1te1Qlrz 4 2 1 NA 1
## 9 R_00zYHdY1te1Qlrz 5 1 1 NA 1
## 10 R_00zYHdY1te1Qlrz 5 2 0 NA 1
## # â„ą 6,390 more rowsTip:
- Use .fill = TRUE for small-sample or subgroup analysis
(helps increase power).
- Use .fill = FALSE (default) when in doubt for safer
estimates.
4. What If Your Data Is Already Clean?
If you already have a clean dataset, use
make_projoint_data():
out4 <- make_projoint_data(
.dataframe = exampleData1_labelled_tibble,
.attribute_vars = c(
"School Quality", "Violent Crime Rate (Vs National Rate)",
"Racial Composition", "Housing Cost",
"Presidential Vote (2020)", "Total Daily Driving Time for Commuting and Errands",
"Type of Place"
),
.id_var = "id",
.task_var = "task",
.profile_var = "profile",
.selected_var = "selected",
.selected_repeated_var = "selected_repeated",
.fill = TRUE
)Preview:
## <projoint_data>
## - data: 6400 rows, 13 columns
## - labels: 24 levels, 4 columns5. Arranging Attribute and Level Labels
To reorder or relabel attributes:
- Save labels:
Edit the CSV (change
order, label columns; leavelevel_iduntouched)Save it as “labels_arranged.csv” or something else.
Reload labels:
Compare using our example:
🏠Home: Home