Wading in the weeds of digital data- Part 2

I mentioned in a prior post about merging various digital data collection files.  Exploring this data has been an interesting experience, particularly prior to heading to the villages where the data is being collected. There are hundreds of ways to look at the data, but the first pass was taking a cursory look at data quality. One common metric that many folks like to report is the how frequently a survey question is answered.

And the missing data here is village name (aka PA) . So we can count all the beans, or have a computer count those beans for us. And we can produce all of these tables and graphs like the one above.  But the core question in my opinion is what that “2” above really means and why.  So in many ways it’s important to walk away from the “data” for a moment and go back to the context. Is the missing village information because…

• the digital data collection form has a glitch?
• Does the question appear unclear to the women who are answering the questions and they are just moving on to the next question?
• Is the data collector chatting with the respondent woman in another village but that village is not on the list and despite reporting on their home village they are unsure of what to check?
• Or is it because they are in a rush and just need to finish the survey and get back to their daily activities so they move quickly to the next question.

Some digital data collection tools can help improve data quality…. to a certain degree. For example, below is a function which requires an answer prior to proceeding to the next survey question. It’s not a full fix (as future posts will discuss) but may help decrease missing data which is just one marker of quality.

A long day before departure.

Today is my last day in Addis before heading to the field. It will likely take us 1/2 day by 4WD to get half way down to the southern tip of Ethiopia. And then another approximately 1/2 day to get to a town of about 20,000 people. Then the next day to one of the two small project sites. To the villages themselves.. another short drive.

Much of yesterday was in the office working on logistics.  Driver, arranging hotel accommodations, sketching out a more detailed travel plan.

• depart early am or mid day?
• stay overnight in Awasa or head straight town and make a full run of it.
• what is the best hotel in the first small town
• is there electricity in the 2nd town, maybe we should stay 45km outside?

I went to the local grocery store to pick up some treats for the road, and will pick up more tomorrow. That in complement with a few Cliff bars should do the trick.

My brain yesterday was slightly fried from doing a preliminary analysis of 298 household surveys, and I couldn’t muster the energy to dive back in. But the timing was good because today I embarked upon a more detailed framework for field interviews, focus group discussions and potential surveys. This may be my only feasible opportunity to print paper surveys so that took top priority. (why not tech? that’s for another post)

From a bird’s eye view it’s a patchwork methodology for capturing lessons learned and measuring the potential early effects of integrating digital data collection into a community-based early warning system. While this is not a research project, there are many research methodologies that are leveraged to achieve the evaluation’s goals. What is remarkable to me is the challenges we face in meshing research approaches with programmatic project goals. For some traditional researchers it looks messy, “lacks rigor”, and even half hazard, I believe this is one area where research meets practice. Or more aptly said in this context where humanitarian practice touch points with research.

A large portion of the field work will  lean upon qualitative methods; semi-structured interviews and focus group discussions. The questions may change based upon who is engaged in the conversation. And the language may also vary from English to Amharic to Afaan Oromoo. We may be in a local government office, at a local restaurant, in a hut, or under the shade of a tree.

From a researcher’s standpoint we can challenge the “validity” or “accuracy” of this approach, but is the purpose broad generalizability of integrating one type of digital data collection into two small areas of southern Ethiopia? Maybe (an likely) that’s not the goal. Maybe the goal is to gain learning/knowledge of how to creatively move forward with a multitude of actors.  So that data and knowledge extraction is not the predominating factor (hopefully) for a manuscript or conference, but information and knowledge sharing for a group of people who may not care at all about peer-review.

First Steps in evaluating ICT – Mixed Fruit & Paper

So I’m back again  in Ethiopia to evaluate a project which has begun to integrate digital data collection into a community-based early warning system. My prior blog here, and here briefly describes the  system which I helped design in 2007 .

I started out the day in Addis scribbling on a piece paper, drinking a mixed fruit juice drink which I’ve greatly missed since my last trip in 2012. The piece of paper is a brainstorm of pictures, arrows and images- rather than any  “data”, “numbers”, columns or even rows of a database. I’m partial to little pieces of paper and scribbling out pictures because it helps me stay closer to the 90/10 rule. In short it’s about the people, processes, partnerships, and trust networks that really push an ICT project forward. That’s the 90%. The remaining 10%  (others would say 20) is the technology itself. I know this brainstorming session will be drastically changed by the insightful and ever powerful conversations with both the local partner and community women. It’s just a start for met to catch up with them.