from your responses I see there is nothing that went wrong with the hosting of the data in the cloud server or with the functioning of the application. Every time you create an assignment - this is an order for an interviewer to deliver a certain portion of the data for your survey. The survey is complete when all the assignments have been completed (either resulting in valid data, or with an acceptable reason why it can’t be delivered, such as “address does not exist” , etc). Verifying that all of the assignments are fulfilled is up to the client, since only the client knows how many observations to collect or when to finish the survey for other reasons (calendar, budget, logistical, etc).
Our server maintenance team makes reserve backups for the data during the period the server is active and they are retained until the server is shut down. They are discarded shortly after the server is shut down and this period has already elapsed in your case.
Thinking about your current problem at hand, provided that the data is sufficiently valuable for your project, the only thing coming to my mind is a workaround, whereas you utilize the data already saved in the tablets as a source of information for the same survey. Basically this means seating a data entry operator for retyping the same data from one tablet to another.
- Request another server from us using the same procedure as for the original server.
- Create one new interviewer account and issue this person a new tablet (or one that has not forgotten to send the data in October, and so can be flushed).
- Collect all the tablets with unsynced valuable data, get the credentials of the original interviewers so that the data entry operator can login to them to see the interview data.
- Re-create assignments as necessary for the data entry operator so that she can enter all of the unsynced interviews. You can use wildcards (leave identifying fields blank) if you trust her that she will enter all the IDs correctly, then 1 assignment for infinite amount should suffice. If you can afford multiple data entry operators split the work between them accordingly.
- let them re-enter the data.
- Synchronize the tablets of the data-entry operators to the new server.
- Export the data.
- Confirm the server can be closed.
- Append (in Stata’s sense) the datasets you exported in October and entered now.
What you retain: validations, logic, closing the non-applicable questions and subsections.
What you loose: GPS, image or audio questions if they were present, paradata will show different timestamps for entries, duration will generally be much shorter, comments will be entered from the name of the interviewer, even if originated from the supervisor.
Provided that you have a non-multimedia questionnaire (no pictures, audio, GPS, etc) you may also use web interviewing mode, so that the data entry operators could utilize any available computers for this task. If your original questionnaire utilized GPS and timestamp questions, your data entry operator clearly will not be able to enter the same original values, so you may go through the questionnaire and change the question types to a set of numeric questions for GPS, and a text for the timestamp. This will also have implications for step Nr9 above, and may have a snowballing effect of changes throughout if there were any logical dependencies/checks in the questionnaire utilizing these questions.
The above procedure is unpleasant and costly. But it is only a result of the interviewers not sending the data, not any defect in the system and is not required under any normal circumstances.
On the other hand if the data is extremely valuable and if it is just a handful of interviews that were not synchronized because the “interviewers forgot”, then it should not be a disaster.
Best, Sergiy Radyakin