ODK notes from Ghana

ODK Community,

I just wanted to send an update from the field on our work in Ghana. Our study, through the Johns Hopkins Center for Water and Health, is a prospective observational study evaluating the impact of community-owned water vending kiosks along with hygiene education on village access to potable water and associated health outcomes. In all about 600 households are interviewed each year in addition to collection of water samples. We switched to using ODK this year from using a paper survey and so far it has worked out really well.

We are about half way done with the data collection and I thought I would take the time to write up some of my comments/experiences with ODK and our hardware. I apologize in advance for the rambling writing but I threw this together pretty quickly.

Hardware:
-We decided to use Mytouch 3Gs and ended up getting a combination of 32As and 32bs with some of the 3.5mm headphone jack editions. In order to display the characters for some of the Ghanaian languages we decided to root the phones and use Cyanogenmod 5.0.8. We rooted the first twenty with no issues and then we ended up bricking a few phones in the process. This only happened with the 3.5mm headphone jack phones and we really don't know what happened. Following the same process as all the other 3.5mm headphone jack phones lead to some phones becoming rooted and others bricked....We decided to cut our losses and go with the rest unrooted.
-To protect the phones from rain and the elements, we used Dry Paks. These are little sleeves that the phones go into with a membrane between the user and the phone. This membrane reduces the sensitivity of the touchscreen but surely protects the device. They also goes around the surveyors neck and make it seem like more official equipment and helps communities identify our surveyors. About half of the surveyors have had no problems using the phones inside the paks while the rest seem to take the phones out during interviews and replace them in between households.
-With the exception of a small team that goes to a remote area for a few days at a time, the phones are returned to a central place each night for charging so no external charging devices were needed (e.g. solios). We have some extra batteries for the small team and have found that the stock batteries last for a day and half of full time use (in airport mode).

Nightly Routine:
-Nightly pulling the data from the phones through script that calls adb. We decided to leave the data on the sd cards as a back up.
-Combining .xmls with KoboPostProcessor. We have played around with local instances of ODK Aggregate and in its current form it is not reliable for our purposes. Some of the issues we have found are that it will write multiple entires if the same file is pushed more than once (it happens with the way our data is stored), and that it sometimes runs out of memory and crashes while forms are being uploaded.
-We have been using Aggregate to plot the points to google earth to help the team understand why they are taking 3 different GPS points per survey. SO far they have been really into seeing the points and ask if their fixes from the previous days were ok.

Training:
We trained 25 surveyors and 5 supervisors , most of whom had already worked on the study in previous years. We spent the first few days going through the survey on paper and nailing down the translations (4 languages including English) of new questions. The updating/tracking of the specifications and translations was made much easier by using google docs. We introduced the devices at the end of day two and went over the basics of touch screens and had them do a short 10 question exercise on ODK. I made the questions pretty funny and the participants seemed to really appreciate it. As suggested by Neil Hendrick, we made great use of the SDK Emulator and even found a great skin to make it look exactly like the Mytouch 3gs we were using. The next day we worked on understanding constraints, logic, changing languages, GPS use and dealing with basic problems like the disappearing (or stuck) keyboard. All this was done with a series of test forms which were gone through first on the big screen then on the individual devices. Of course, once the devices were handed out it was hard to have anyones attention at the front of the room.

The rest of the training was spent working with the actual survey and going over trouble shooting. Like the other forms we went through it on the big screen first, then moved to individual interviews and finally had people interviewing each other. Since people had to do this in multiple languages and since the survey is about an hour long, it took a while to get through all of this. During the training we found that the surveys crashed due to memory issues every once and a while. To overcome this we split the translations up between surveys so that there was only English plus one Ghanaian language on any form. The surveyors would have to choose the appropriate language before they use the phone.

The surveyors picked up the use of the phones pretty quickly. Most of them were young and quite tech savy so the jump from computer and basic cell phone use to the smart phones was pretty simple. It did however take a bit of time for them to get the hang of dealing with text input fields due to the size of the keys and the challenges in editing text already typed in. The surveyors also got really frustrated with the length of time it take to get the first GPS fixes. After 3 or 4 times at it, the satellites were in the almanac and they got the hang of it.

Random Comments:
-Many studies like ours call for random back checks of surveys. We decided not to teach the team how to go back to surveys to view answers for fear of them changing an answer on accident. This may have been a mistake and I plan on revisiting this for the next survey. Instead we developed supervisor surveys with a subset of questions from the fill survey that supervisors use to go back to the household. So far this has worked out well.
-As Neil suggested, I have the surveyors do a morning exercise to warm up and get a solid GPS fix in the survey area.

We plan to continue using ODK for our research and will hopefully build our own edition in the coming year. Our hopes for future ODK releases (some of which I know will be in the next versions) include:

Collect:
-Support for different user types (surveyor, supervisor, admin)
-Support for multiple images in questions
-Improve memory issues so that long surveys with multiple languages don't crash
-Ability to have order of selections in select or select1's randomly changed to reduce bias

Aggregate:
-Improved support for local instances
-Data security/ encryption (the current version would not pass our Internal Review Board)
-Better support for submitting forms in areas with unreliable cell networks

Regards from Accra,

andrew

Hii Andrew,
i am Llord Titus from the Volta region of Ghana.
Actually I am into enviromental advocacy and the ODK is a very needed
tool for my activities.
How can I get one?
My e-mail is titusllord@gmail.com (remember the l in is double)
Thanks

ยทยทยท On Aug 16, 6:20 pm, Andrew Azman wrote: > ODK Community, > > I just wanted to send an update from the field on our work in Ghana. Our study, through the Johns Hopkins Center for Water and Health, is a prospective observational study evaluating the impact of community-owned water vending kiosks along with hygiene education on village access to potable water and associated health outcomes. In all about 600 households are interviewed each year in addition to collection of water samples. We switched to using ODK this year from using a paper survey and so far it has worked out really well. > > We are about half way done with the data collection and I thought I would take the time to write up some of my comments/experiences with ODK and our hardware. I apologize in advance for the rambling writing but I threw this together pretty quickly. > > Hardware: > -We decided to use Mytouch 3Gs and ended up getting a combination of 32As and 32bs with some of the 3.5mm headphone jack editions. In order to display the characters for some of the Ghanaian languages we decided to root the phones and use Cyanogenmod 5.0.8. We rooted the first twenty with no issues and then we ended up bricking a few phones in the process. This only happened with the 3.5mm headphone jack phones and we really don't know what happened. Following the same process as all the other 3.5mm headphone jack phones lead to some phones becoming rooted and others bricked....We decided to cut our losses and go with the rest unrooted. > -To protect the phones from rain and the elements, we used Dry Paks. These are little sleeves that the phones go into with a membrane between the user and the phone. This membrane reduces the sensitivity of the touchscreen but surely protects the device. They also goes around the surveyors neck and make it seem like more official equipment and helps communities identify our surveyors. About half of the surveyors have had no problems using the phones inside the paks while the rest seem to take the phones out during interviews and replace them in between households. > -With the exception of a small team that goes to a remote area for a few days at a time, the phones are returned to a central place each night for charging so no external charging devices were needed (e.g. solios). We have some extra batteries for the small team and have found that the stock batteries last for a day and half of full time use (in airport mode). > > Nightly Routine: > -Nightly pulling the data from the phones through script that calls adb. We decided to leave the data on the sd cards as a back up. > -Combining .xmls with KoboPostProcessor. We have played around with local instances of ODK Aggregate and in its current form it is not reliable for our purposes. Some of the issues we have found are that it will write multiple entires if the same file is pushed more than once (it happens with the way our data is stored), and that it sometimes runs out of memory and crashes while forms are being uploaded. > -We have been using Aggregate to plot the points to google earth to help the team understand why they are taking 3 different GPS points per survey. SO far they have been really into seeing the points and ask if their fixes from the previous days were ok. > > Training: > We trained 25 surveyors and 5 supervisors , most of whom had already worked on the study in previous years. We spent the first few days going through the survey on paper and nailing down the translations (4 languages including English) of new questions. The updating/tracking of the specifications and translations was made much easier by using google docs. We introduced the devices at the end of day two and went over the basics of touch screens and had them do a short 10 question exercise on ODK. I made the questions pretty funny and the participants seemed to really appreciate it. As suggested by Neil Hendrick, we made great use of the SDK Emulator and even found a great skin to make it look exactly like the Mytouch 3gs we were using. The next day we worked on understanding constraints, logic, changing languages, GPS use and dealing with basic problems like the disappearing (or stuck) keyboard. All this was done with a series of test forms which were gone through first on the big screen then on the individual devices. Of course, once the devices were handed out it was hard to have anyones attention at the front of the room. > > The rest of the training was spent working with the actual survey and going over trouble shooting. Like the other forms we went through it on the big screen first, then moved to individual interviews and finally had people interviewing each other. Since people had to do this in multiple languages and since the survey is about an hour long, it took a while to get through all of this. During the training we found that the surveys crashed due to memory issues every once and a while. To overcome this we split the translations up between surveys so that there was only English plus one Ghanaian language on any form. The surveyors would have to choose the appropriate language before they use the phone. > > The surveyors picked up the use of the phones pretty quickly. Most of them were young and quite tech savy so the jump from computer and basic cell phone use to the smart phones was pretty simple. It did however take a bit of time for them to get the hang of dealing with text input fields due to the size of the keys and the challenges in editing text already typed in. The surveyors also got really frustrated with the length of time it take to get the first GPS fixes. After 3 or 4 times at it, the satellites were in the almanac and they got the hang of it. > > Random Comments: > -Many studies like ours call for random back checks of surveys. We decided not to teach the team how to go back to surveys to view answers for fear of them changing an answer on accident. This may have been a mistake and I plan on revisiting this for the next survey. Instead we developed supervisor surveys with a subset of questions from the fill survey that supervisors use to go back to the household. So far this has worked out well. > -As Neil suggested, I have the surveyors do a morning exercise to warm up and get a solid GPS fix in the survey area. > > We plan to continue using ODK for our research and will hopefully build our own edition in the coming year. Our hopes for future ODK releases (some of which I know will be in the next versions) include: > > Collect: > -Support for different user types (surveyor, supervisor, admin) > -Support for multiple images in questions > -Improve memory issues so that long surveys with multiple languages don't crash > -Ability to have order of selections in select or select1's randomly changed to reduce bias > > Aggregate: > -Improved support for local instances > -Data security/ encryption (the current version would not pass our Internal Review Board) > -Better support for submitting forms in areas with unreliable cell networks > > Regards from Accra, > > andrew