Sign in
x
☰

Condense text open-ends

Squish multiple responses into one distribution

Updated at 2019-06-26 14:41:38 UTC

Protobi has a feature to condense multiple columns of text open-end data into a single distribution. This is useful when the survey provided multiple text-boxes to enter open-ends for one question, or the survey asked a series of open-end questions and you would like to analyze the responses together. 

For example, the question below asks about three different products.  For analysis we'd like to combine these three open-ends into one single list.

Protobi interface showing parent group qV with three text open-end questions (qQ2, qQ3, qQ5) displayed side-by-side. Each sub-question asks "What is the primary reason for the score you assigned to" followed by different products (A, B, C). Each column displays text responses with percentages, including responses like "Reputation", "Experience", "Good reputation", "Not familiar", etc., with most showing 1.1-3.3% frequency. Filter boxes and Apply buttons appear at the top of each column.

To condense the values, group the three questions under one parent qV. Press the edit icon for the group and select "Transform..." from the context menu to bring up the transform dialog:

Transform group dialog box with blue header showing radio button options for calculating values. Options include None, Sum, Mean, Median, Min, Max, Std Dev, Condense ("squish" with horizontal arrow, selected with blue filled radio button), and Compact ("squash" with up arrow). A "Hide children" checkbox appears at bottom, unchecked. Cancel and Ok buttons in gray and blue respectively at bottom right.

The text responses are now condensed into a single distribution.   

Protobi view of condensed parent group qV showing combined text responses in a single column. The header reads "What is the primary reason for the score you assigned to ...?" with italic text "Transform to squish" below. A Filter box and Apply button appear at top. The distribution shows "Experience" at 4.3%, "Reputation" at 3.3%, "Not familiar" at 3.3%, followed by numerous other responses at 2.2% or 1.1%, including "More experience", "personal experience", "Good company", "Good reputation", "Good results", and many others with truncated text indicated by ellipses.

This is a multi-value element, so percentages may sum to more than 100%. For instance, people who chose "Experience" also entered other reasons:

Protobi view of condensed qV showing filtered results highlighting respondents who mentioned "Experience". The top shows "Experience" at 100.0% with a long yellow horizontal bar. Below are nine other responses each at 25.0% with shorter yellow bars: "I dont know the rep", "Lack of exposure", "The rep is well known to me", "Knowledge", "Use it more", "Used it in the past", "Facts", "Good use", and "Reputation". This demonstrates the multi-value nature where one person provided multiple text responses.

Counting unique respondents vs unique responses

By default the every element uses the number of respondents as the denominator for frequencies. In a condensed open-end group, one respondent likely has more than one response so frequencies will sum to >100%

In the JSON editor you can add an attribute "unique" and set it to false to show percent out of unique responses instead. This will result in a distribution that sums to 100%.

JSON editor dialog showing Element properties for qV. The JSON code displays configuration including key "qV", title, type "string", chartOptions object, children array containing "qQ2", "qQ3", "qQ4", and on line 25 shows the attribute "unique":false highlighted in gray, followed by other properties like displayKey, maxValue, transform "squish", and hideChildren set to true. Cancel and Ok buttons appear in top right.

Below, Q22 is showing percent out of unique responses. Q22_copy shows the default which is percent out of unique respondents. 

If we filter the view on one respondent, you can see in Q22 N is the number of response values. In Q22_copy we count unique respondents, so Protobi recognizes the different responses are from one respondent and N is 1. 

Side-by-side comparison of three Protobi elements: Q22 (left), Q22_copy (center), and respId (right). The view is filtered to one respondent. Q22 shows "transform to squish" with three responses ("Environmentally friendly impor...", "Well eytablished studies", "Easy to use") each at 33.3% with blue bars and N=3. Q22_copy also shows "transform to squish" with the same three responses but each at 100.0% with blue bars and N=1. The respId column on right shows filter applied (50625) at 100.0% with yellow bar and additional respId numbers below at 0.0%.

Remove the transform

To undo a transform, go to the group's context menu and select "Expand/collapse." Then, bring up the transform dialog and select "None." The group should return to its original state.

Video tutorial

Condense multiple text columns

Condense text open-ends video

See related

Transform dialog