Hi, I am working with Network Canvas Interviewer on Android tablets and have noticed that occasionally when sessions are exported, the resulting CSV files are blank or only have the header rows. This happened once when I was training our interviewers and I simply exported the data again and everything worked, so I did not worry much about it. However, we did pretesting yesterday and the issue has recurred for several of the interviews we conducted. I don’t have access to check the interviews themselves at the moment, but I know the first time this happened during training everything looked correct in the interview session. Could someone look into this issue? I don’t want my enumerators to have to repeatedly export their data and check that it is correct during their field work.
Hi!
I’m really sorry to hear about this. Definitely not what you want as you enter the field.
Can you just confirm a few things?
- You said that the exact same sessions sometimes export and sometimes don’t. Is that the case? In other words, there are no sessions that consistently don’t export…?
- Which specific android tablet are you using, and which version of Android?
- Can you describe in as much detail as possible what your export workflow is for your study? For example, how many interviews are you exporting at a time, and which export options are selected?
- Do you have a timeline for when you are starting data collection? It is important that I manage expectations about fixing this, but I do want to do my best to get to the bottom of this for you so that you can complete your work.
Finally, if it is possible, I’d appreciate it if you could share your protocol with me so that I can test if it the issue is caused by something specific to it. You can email it to joshua@northwestern.edu.
Hi Joshua, thanks for your quick reply.
- This is correct, I have been able to eventually get all interviews to export
- We are using Galaxy Tab A7 lites. I’m not sure if all the tabs are on the same Android version, but I know this at least occurs on version 14 on my personal tab.
- We aren’t exporting many sessions; yesterday this was happening with two interviews per enumerator. In the fiels, tbey will be doing this daily, so that load may be higher during collection.
- We start data collection in about a week, though if it takes a bit longer to fix this it shouldn’t disrupt our workflow too much.
I can indeed share the protocol with you, I will try to within the next 24 hours. Thank you again!
Thanks for getting back to me.
Could you just confirm which export options you have selected?
Yes sorry, we are exporting only as CSV—no other options are selected.
Is it any particular file that is consistently blank/headers only? For example, the ego file, one of the two edge lists, or the person type node attribute list?
Can you share how your cases are named?
Also, you mentioned you encountered this issue when training and pretesting - were you entering data on all screens during this, or only specific screens? If it was specific screens, can you say which ones? Were you entering plausible values, or just random characters?
So far, I am not having any luck reproducing the issue in the simulator, using dummy data or manually entered data.
This was done with full test interviews conducted in the field, so all required info (non skippable questions) and most skippable questions were answered. We are naming cases with unique IDs, a letter followed by some numbers.
This seems to happen most frequently with the person attribute list, but has happened with other files as well.
Could it potentially be that the enumerators are moving off the Network Canvas app before it is done exporting? I don’t think they are closing it, but maybe they need to leave it on the screen for longer? I don’t think there is a consistent message that says when the upload is done though (we upload to both a cloud server and internal memory).
Once the progress bar is complete and the share menu is triggered, Interviewer has finished encoding the files. If someone were to open another app while the encoding is still happening, I suppose that could perhaps cause an issue such as this.
However, since you said you encountered the problem with as few as two sessions (which should take a couple of seconds at most to encode) that seems unlikely to be the cause.
I’ve still not been able to reproduce, sadly. I’ve been exporting to google drive and CX File Explorer.
If you have time, could you try generating dummy sessions (Settings → Developer Options → Generate Test Sessions) and seeing if those trigger the problem when exporting? If they do, that would probably confirm this is an issue with the physical hardware, and I will switch to trying to test on a real device.
I tried doing the dummy cases and creating my own and didn’t encounter the issue again. I will let you know if the issue recurs and if so will try to send you as much information as I can about the interview for which it happened.
Okay, that’s reassuring in some ways.
We had a pretty severe bug in December last year regarding Android and CSV export (see here: Important Information Regarding a Recent Software Bug in Interviewer). I am wondering if the fixes we implemented there might have indirectly caused this bug.
I have attached the version of Interviewer prior to that bug being fixed. I would be interested to know if you are able to trigger the issue using this version. Please remember that this version has a known serious bug in the exporter code, which is triggered by using the merge sessions option and repeatedly exporting the same data. Please only use this version of the app as a diagnostic tool to help us find the cause.
Hi Joshua, I installed the version of NC interviewer you attached and was able to replicate the issue there as well. See the screenshot of one of my sociogram stages from a dummy interview I generated and the corresponding output edge list file, which only has the header row. In this case, though, all the node attribute files were filled.
Just confirming that this also happened again with the new version of the app–this time the edge list for one of my sociograms was blank, but all the other files were filled (like in the picture in the comment above). I am also not able to fix it this time–I have tried exporting again several times now and the same issue arises. I have also checked and the interview definitely has connections for that sociogram stage.