Usecases
Different use-case scenarios
This page lists a number of use case scenarios and suggestions on how to solve problems with SDAPS. SDAPS is highly flexible and it may not always be obvious how to best solve a task.
In general, you should consider the risk of errors and their impact.
Anonymous survey with up to two pages
This is the best case scenario. One piece of paper (double sided) contains one questionnaire. There is no risk for pages from different subjects to be mixed. Data entry errors will only occur due to e.g. bad check marks or problems with the scan.
Anonymous survey with more pages
In multi-page scenarios you need to consider the impact of pages getting mixed up. How to handle this may depend on the size of the survey.
For smaller scope surveys it can be sufficient to ensure that pages are not mixed up. This is an error prone method and is only feasible for small scale survey with a well controlled environment. Be careful, a single page getting lost or mixed up during scanning can cause ripple effects and make data extraction from large parts of the survey impossible!
Background
SDAPS will usually assume that the pages for one questionnaire are together in the scan. This group of pages is internally called a “sheet” and forms a single row in the CSV export. The order of the scan within this group does not matter, so a scan like the following is fine:
A1, A2, A3, A4, B4, B3, B2, B1
Here the first four pages belong to subject A and the next four to subject B. This works fine, but imagine that pages are mixed up, for example of on the following scenarios:
A1, A2, B3, B4, A3, A4, B1, B2
A1, A2, B1, B2, A3, A4, B3, B4
In this case, SDAPS will group the first 4 pages and you have data from both subjects in each of the two exported rows. Even worse, in the second example SDAPS cannot even completely export the data as pages are duplicate.
Solution
SDAPS currently only provides one proper solution to this problem. The idea is that the questionnaire handed to each subject can be uniquely identified. We assign a pseudonym to the subject and print it as a barcode/QR-code onto each page. Then, when pages get mixed up we know what belongs together and can “reorder” the dataset to correct it.
This means:
- Set
print_questionnaire_id
in the LaTeX documentclass header - Run
sdaps setup tex
as usual - Run
sdaps stamp
to generate unique questionnaires that can be printed - Run
sdaps add
to add the scans - Run
sdaps recognize -i
this is faster than a full recognition - Run
sdaps reorder
so that the internal grouping in SDAPS is correct - Run
sdaps recognize
for the data
Note that you can pass the -f
option to the add
command in case the page
count is already wrong. That means you are already missing pages or have
unrelated excessive pages.
Alternatives
There are currently no alternatives. That said, SDAPS could be modified to only process one double page at a time. Then each row in the exported CSV would only represent a partial questionnaire and custom methods can be used to aggregate the data again.
Such a solution would only make sense a method other than the SDAPS provided “questionnaire ID” must be used.