How to Address Changes in the Questionnaire in a Tracker

From Q
Jump to navigation Jump to search

From time-to-time, questions in a tracker will change. Most commonly, this will involve the addition or deletion of categories (e.g., adding new brands), although more fundamental changes may occur (e.g., turning a single response into a multiple response question).

Accidental changes when outfiling data

Sometimes data from a tracker are exported inconsistently. For example, in one wave gender may be in a variable called gnd and in another data file it may be called q2. When you import the updated data, Q will automatically identify such problems, provided that you are not using a CSV file. Although such problems can usually be fixed within Q, it is almost always preferable to have the person that provided that data file re-export so that it is consistent with the original data file structure.

Ad hoc changes

Common ad hoc changes include adding additional brands and changing image statements.

New codes in Pick One and Pick – One Multi questions

Where a new code is added to a Pick One or Pick One – Multi question, this is almost entirely automatically handled by Q. Where you are required to make decisions about how to deal with the change, these are flagged both at the time of import and in the Status column in the Variables and Questions tab. Please note, however, that any analyses done on the variables will implicitly assume that same code frame has been employed throughout the data. That is, if a new brand is introduced in 2010, any analyses run in 2009 will show the brand (but with a value of 0 or NaN, depending upon the analysis). A solution to this is to export the data as Pick Any questions, as then missing values can be used in periods where the data did not exist.

Removed codes in Pick One and Pick – One Multi questions

If a code is deleted in the questionnaire (e.g., a brand has left the market and is not asked in the questionnaire), if there is a need to continue to report on the brand, then the code should still be included in the exported data file.

Changed code frames in multiple response questions

When multiple response questions are changed, such as adding a new brand, this can cause considerable complexity, particularly if the question is then used to construct filters and other questions. In general, the best approach to dealing with this is to anticipate this possibility when setting up the data collection program (see Data File Setup for Tracking Studies ).

The various tools for Setting Up Data are useful when dealing with questionnaire changes. Where there are two different versions of a question, these can be automatically combined in the Variables and Questions tab using Insert Ready-Made Formula(s) and Merge Questions.

When a new brand is added to a multiple response question and there are Binary - Complicated Filters that rely on the existing question then it is more difficult to add the new variable to the question. This is because the complicated filters rely on the structure of the question, and as a result you will not be allowed to use Set Question. Note that this problem does not occur for JavaScript variables. One way to deal with this problem is to take a copy of the question before making changes. The binary filter can then be modified to refer to the new copy of the question. This can become impractical if there are a large number of binary filters that refer to the question. In this case, you can use the technique outlined in How to Add New Variables to Questions used by Complicated Filters.

See also