Hello!
I would like to add a critical rule, with the “block” action, so that an interviewer would not be able to send interviews with errors (i.e. where at least one validation condition is not met). I thought I could use the system generated variable “has__errors” with the condition has__errors > 0 but this does not work. Anyone has an idea of how I can sort this out? Thank you in advance!
Most (if not all) of the system-generated variables are not accessible from the questionnaire syntax. These include has__errors and other properties of the interviews like interview__id or interview__key, for example.
Thank you for your feedback! I wish this was possible or at least there were a similar solution.
Can you tell more about which survey/project needs this?
This is something I would like to use in all my surveys to improve the quality of the data, especially in an upcoming schools’ census in CAR.
Could you please tell a bit more about this census? Who is conducting it? When? Questionnaire? Stakeholders?
Dear @sergiy ,
Sorry for my late response. I was busy with the data collection operation, which is now almost completed. I ended up using susoapi to reject interviews with errors or unanswered questions, but this wasn’t easy because of the Internet issues in the country. The client is the ministry of national education (MEN) of the Central African Republic (CAR), with a World Bank financing through a project. The census is carried out by CERPEPP Afrique, a survey firm based in DRC, to which I bring my expertise as a Consultant (CAPI Expert, from Zechlab, a survey firm based in Abidjan, Côte d’Ivoire). The operation started early November and we are now processing with the data cleaning. We have collected data from all pre-, primary and secondary schools all over the country (whereever possible). We have also collected informations on the ministry staff, including all decentralized entities all over the country. I hope this answers your questions!
Thank you for your valuable and constant support!
There are several packages named susoapi: one for Stata; one for R. If the susoapi package is mine (here), please feel free to create one or more issues on the repo to describe any challenges faced or propose any improvements. (Sadly, I can’t do anything about the internet.)
Sounds like in that case you don’t need validation at all — Just write your checks once as critical rules. The difference is of course that the interviewer will see the error messages not immediately, but by the time she accesses the Complete screen, but if this census has just a minimal questionnaire, then it may be worthy.
Keep in mind also that the errors may also involve supervisors questions, and so not being able to correct the error may create a deadlock for an interviewer. So check what you are restricting.