Restarts and snapbacks with least full quotas
Least Full quota control takes a set multiple-choice answers and pends only the least full cell related to those answers. You may choose how many cells to pend. For example, you may allow respondents to choose any number of responses but will always pend only the two cells containing the fewest number of respondents. Since the least full cells will vary as the survey progresses, it is inevitable that the cells that are available to be pended when an interview starts and unlikely to be the cells that are available for pending if the interview is stopped and then restarted. Indeed, in very busy projects, the least full cells may change between the time the question is first answered and the time the interviewer or respondent snaps back to check or change that answer.
Quota control and snapbacks and
Quota control in restarted interviews explain the general rules and behavior in these situations and highlight the points at which extra care is required. In particular, they note the times when you may need additional scripting to ensure consistency between the quota cells and the interview data.
If you are using least-full pending, you may need to take even more account of this behavior in your script because changes in the cells pended may have far reaching effects on your data. To understand the full extent of what’s involved, take the previous example but, this time, assume that only the least full brand bought is pended.
When the respondent starts the interview and chooses brands A, B, and C, the cell for brand A is pended as being least full and the respondents answers subsequent questions about brand A only. When the respondent restarts the interview, brand B is now the least full so that cell is pended instead of the cell for brand A. The interviewing continues replaying the interview from the case data because the answers are still valid responses for the questions that it needs to ask. It does not know that the answers refer to brand A rather than brand B. The interviewing program reaches an unanswered question and displays it. If the question is filtered based on the cells pended and the question text includes the brand name, the respondent will answer the question based on their experience of brand B. So, not only will the final case data contain out-of-date answers, it will also contain data related to a mixture of brands.
Use the following questions to help you make decisions about how to deal with this.
▪Do you want to pend the quota cells that were least full when the interview originally started or would you rather pend based on the current counts? If you want to repend the original cells, what if some cells are now full? Will you continue with the ones that are still unfilled even if this means discarding some responses that would originally have been used?
▪If all the originally pended cells are now full, do you want to pend a new set of cells or would you rather terminate the interview?
▪What if the respondent has snapped back to the quota controlled question and completely changed the answer so that none of the originally pended responses are chosen?
▪If other cells have become available for pending since the original quota test (other interviews with pended counts may have been terminated so that the pended counts were rolled back), do you want to see whether those cells can be pended for this interview?
▪If the cells that are pended change and there are later questions that are based on the pended answers, how are you going to ensure that the data for those questions is correct?
The following topics suggest ways of scripting to answer these questions:
For a complete example that brings together all these decisions in a single script, see the LeastFullQuota.mdd example script that is installed in [INSTALL_FOLDER]\IBM\SPSS\DataCollection\7\DDL\Scripts\Interview\Quota\LeastFull. Note that this script is an example of how to implement least full pending: it does not illustrate the default method of quota control for multiple-choice response lists.
Note In all cases, saving the names of the pended cells as the responses to a question makes restart scripting much easier because you always have this question available for comparisons between the original and restarted versions of the interview.
See also