Displayr doesn't have a direct connection to QuestionPro, so you import QuestionPro data by exporting a file from QuestionPro and uploading it. QuestionPro's SPSS (.sav) files don't follow Displayr's recommended SPSS file specifications, and QuestionPro offers limited export settings that change this. This article lists the known problems and possible fixes with QuestionPro data and how to fix each one in Displayr.
If you get a poor result that isn't covered here, contact us at support@displayr.com.
Before you field your survey
These steps prevent several of the problems below:
Finalize the questionnaire before fieldwork starts. Adding a question (even a hidden one) partway through can renumber the variables after it. Displayr links your tables to variable names, so renamed variables break existing outputs when you update the data.
Use a standard Ranking question, not Drag and Drop. Also add validation so respondents can only rank the number of items you ask for.
Keep answer option text short. Question text plus option text over 255 characters gets cut off.
For MaxDiff or conjoint, check that QuestionPro can export the experimental design. See MaxDiff and conjoint data below.
Multiple-response questions show 100% for every option
What you'll see: Every option in a multiple-response (Binary - Multi) table shows 100%.
Why: QuestionPro records 1 for selected and leaves not-selected options blank. It never records 0. Displayr reads the blanks as missing data, so the only people counted for each option are the ones who selected it. This also means there's no way to tell "saw the option and didn't select it" apart from "was never shown the question".
How to fix it:
Select the variable set in the Data Sources tree. If it has a caution icon, you can click Fix instead and go to Step 3.
In Properties
, click Data > Attributes > Values.
Change Missing Values to Include in analyses for the blank/missing code, and make sure the selected code (1) is counted.
Click OK.
OPTIONAL: If only some respondents were asked the question, rebase it to people who gave at least one answer, select the variable set, and click + > Data Quality > Rebase Multiple Response Data in Variable(s) to NET. See How to Rebase Multiple Response Data in Variable(s) to NET.
If the people who weren't asked also left no answer, rebasing to the NET removes them correctly. It also removes anyone who was asked but selected nothing. If that matters, filter the question by the question that controlled who saw it instead. See How to Rebase One Question Based on Another Question.
Multiple-response options appear as separate questions
What you'll see: Each option of a "select all that apply" question appears as its own variable set, or the options import as Nominal instead of Binary - Multi.
Why: QuestionPro puts each option's text in the value label for code 1 (for example, 1 = "Piano"), instead of using a common value label like "Selected" and putting the option text in the variable label. Displayr uses matching labels to recognize multiple-response questions, so the options aren't grouped.
How to fix it:
In the Data Sources tree, select all the variables for the question.
Right-click and select Combine.
Ensure that the Structure is set to Binary - Multi in Properties
In Data > Attributes > Values, ensure that Count this Value is ticked for code 1 (Selected, Yes, etc.). See How to Set Value Attributes for a Binary-Multi and Binary-Grid.
Click OK.
If the category names disappear after combining, copy each option's text from its value label into its variable label via right-click > Rename.
After combining, you may also need the fix in Multiple-response questions show 100% for every option.
Labels are cut off or blank
What you'll see: Row or column labels in tables are missing, cut off partway, or show the question text but no statement.
Why: QuestionPro limits variable labels to 255 characters, question text included. With long question wording, the label often ends before the option or statement text, so Displayr has nothing to use as the category name.
How to fix it:
Select the affected variables in the Data Sources tree.
Right-click and select Copy Labels.
Paste into a text editor, add the missing option or statement text from your questionnaire, and copy it again.
Right-click the affected variables in the Data Sources tree and select Paste Labels.
To avoid this in future surveys, keep question text short, or move long instructions out of the question text in QuestionPro.
Variables import as text
What you'll see: Most or all variables show as Text in the Data Sources tree and produce lists of responses instead of tables.
Why: Some QuestionPro .sav exports store every variable as a String (text) type. Displayr keeps the variable types in an SPSS file, so text variables stay as text.
How to fix it: Select each affected variable and change its Data > Attributes > Structure to the correct type, such as Nominal, Numeric, Date/Time in the Properties panel.
Answer options no one chose are missing
What you'll see: A question shows fewer categories than your questionnaire, for example, a 1–5 scale with no value of 1. You may also see a "fundamentally different structures" warning when combining variables.
Why: QuestionPro exports only the answer options that at least one respondent chose, instead of the full list of options. Variables from the same question can end up with different categories, and Displayr won't group them.
How to fix it: Add the missing categories back. See How to Add Empty Categories to a Question. Then combine the variables as described in Multiple-response options appear as separate questions.
Unanswered questions show as 0
What you'll see: A "0" category appears in tables for respondents who didn't answer.
Why: Some QuestionPro exports code unanswered questions as 0 instead of leaving them blank.
How to fix it:
Select the variable set in the Data Sources tree.
In Properties, click Data > Attributes > Values.
Change Missing Values for code 0 (Missing data) to Exclude from analyses.
Click OK.
Only do this if 0 isn't a real answer option for that question.
Tables break after updating the data
What you'll see: After updating your data set with a new QuestionPro export, you get many data difference warnings, and some tables or banners go missing or show the wrong question.
Why: If a question was added or moved in QuestionPro during fieldwork, QuestionPro renumbers the variables after it (for example, Q16 becomes Q17). Displayr matches variables by name, so it can no longer find the originals.
How to fix it: No setting in Displayr prevents this. Rebuild the affected outputs using the new variable names. To avoid it, finalize your questionnaire before fieldwork starts.
Ranking questions don't show a table of items by rank
What you'll see: A ranking question imports as variables named like Q22Rank1, Q22Rank2, and so on. Each one holds the code of the item ranked in that position, not the rank each item received. Some respondents may have ranked more items than you asked for.
Why: QuestionPro stores ranking data by rank position. Drag and Drop questions also don't enforce a "rank your top N" limit unless you add validation.
How to fix it:
-
For each item, create a numeric JavaScript variable that returns the rank the item received. For example, for item 1:
if (Q22Rank1 == 1) 1; else if (Q22Rank2 == 1) 2; else if (Q22Rank3 == 1) 3; else NaN Select the new variables, right-click, and select Combine.
Ensure the Data > Attributes > Structure is set to Nominal - Multi. You'll get items down the side and ranks across the top.
If some respondents ranked more than your limit, only reference the rank positions you asked for (for example, Rank1 to Rank3). Consider reporting the percentage who ranked each item in their top N rather than an average rank.
See How to Analyze Ranking Data.
Grid questions aren't recognized
What you'll see: A grid question imports as separate variables, or changing its structure to a grid has no effect.
Why: QuestionPro sometimes exports the rows of a grid with different values or labels, so Displayr doesn't see them as one grid.
How to fix it: Select the variables, right-click, and select Combine. Ensure the Data > Attributes > Structure in Properties is set to the grid structure you need. See How to Combine Variable Sets as a Grid.
Looped questions have several rows per respondent
What you'll see: The data file has more rows than respondents, and looped questions can't be crossed with other questions.
Why: QuestionPro can export looped questions in a stacked (long) layout, with one row per loop instead of one row per respondent.
How to fix it: Ask QuestionPro for a file with one row per respondent. If that's not possible, unstack the data. See How to Unstack Data Using R.
Duplicate responses
What you'll see: The same respondent appears more than once.
Why: QuestionPro can record a new response when a respondent uses the browser's Back button instead of the survey's Back button.
How to fix it: Remove the duplicate cases from your data file, or ask QuestionPro to remove them before you export.
MaxDiff and conjoint data
QuestionPro's MaxDiff and conjoint exports usually aren't in the layout Displayr needs, and you'll need to restructure the data before you can analyze it. For MaxDiff, Displayr needs:
An experimental design showing which alternatives each respondent saw in each task (version, task, and alternative columns).
Respondent data with one row per respondent and one best and one worst variable per task, each coded with the alternative number.
Known problems:
No experimental design. QuestionPro randomizes alternatives and often doesn't export a design. Without one, you can't run Hierarchical Bayes or latent class analysis, but you can still run a counts analysis. See How to Perform Counts Analysis on MaxDiff Data. If QuestionPro can export which alternatives each respondent saw, you can rebuild the design from that.
Responses stored one variable per alternative, or stacked. Restructure them into one best and one worst variable per task, with one row per respondent.
Binary Direct (anchored) MaxDiff uses different codes. QuestionPro codes 0 = least, 1 = not selected, 2 = most. Displayr's anchored MaxDiff expects -1 = least, 0 = not selected, 1 = most, so subtract 1 from each value.
Conjoint exports of level counts. If QuestionPro exports the number of times each level was chosen, the data can't be converted for choice modeling. You need an export with one row per respondent per task, showing which option was chosen.
Before fielding, ask QuestionPro to confirm they can provide the design and respondent data in these layouts. See How to Perform an Advanced Analysis of Experimental Data (MaxDiff).