Letting the data speak (harder than it sounds)

This week I have been “playing” with my data and trying to listen. I want to tell their story, and not impose what I think they should be telling me (This was a point brought up in Writing Science, and when doing exercises in the book it was clear that articles that force stories are not a good read and you can’t cite them). I think the particular challenge for my data set is that depending on what set of assumptions I use to calculate the P flows in the city food system or the UA system really influences the results. Of course I am being very transparent about assumptions and calculations and my reasons for using them, but now I need to make the tough judgement call about which option to chose.

I am also struggling with which stories to tell. Going back to the research questions helps, but it isn’t clear yet if I should write about two stories, or one big one, and in what order sub-stories should then be told. Hopefully Christmas magic will help me create a clear story line with good text, good numbers, and amazing figures!IMG_3731

As I explore my data and try to write an academic story, I am also trying to decide what short story I should share with all the stakeholders that contributed to this research. I had started this process earlier in the season, but now with real numbers coming in, I need to make a call about what the take-home messages should be.

Thinking about going back to basics, explaining the problem and what sustainability thinking and my data have to say about it has also made me reflect on sustainability solutions in general. As sustainability scientists, and systems’ thinkers, we can offer policy, technological, behavioral, cultural, and political changes that can help increase sustainability; but ultimately, the “highest” level of solutions, often come from or with education. That is, increasing knowledge but even more importantly ways of thinking and understanding the world around us. Even though “we” acknowledge the importance of learning and education, sometimes learning/teaching can be the solution that feels most out of reach because changing education and reaping the benefits take a long time. I personally don’t know where to start that change a lot of the time. One of my friends shared a link to one of his friends company who develops educational toys called twenty one toys and I must say I felt pretty good about starting there this christmas.

Processing data

data processingOver the past two weeks I have been taking my surveys and “transforming” the answers into absence and presence data and quantitative P flows. I had already done this process for about 50 or 60 of the surveys in july in order to check if the information was complete and how I should be re-checking the other surveys after I had completed them to make sure everything I needed was there.

I am so happy I took the time to do this “test” processing and took lots and LOTS OF NOTES and made good summary excel sheets of all the conversion factors I had used. It is definitely making my job easier now. I am still finding some “novel” conversions that I need to look-up and I am also thinking hard about the conversions I used before to make sure I am ok with the assumptions behind them. Still, the task does seem a little less overwhelming as I did the prep-work a few months ago.

I chose to start the processing even though I am still waiting for some missing pieces of information from some stakeholders. This is making the processing a little more tricky because it means I need to go back in complete the surveys and make sure those changes make it all the way through even when I change excel files (I am actually not a very big fan of linking between excel files because I have had too many problems when changing computers and to me it makes the data less sharable with people). I am using COLOR CODES for the cells that need extra information and I think this should be enough to keep me on track.

Here are my basic steps from going from the filled-out surveys stored in limesurvey to something I can use:

  1. Go back to research questions. What are you looking for (but there is a certain level of change, like my questions are quantitative but after the pilot I really also did some yes/no questions to be sure I wouldn’t loose possible data and also “ease” people in)
  2. What steps need to be taken between what I have and what I need. So for me its to convert everything to P, so unit conversion (all metric, but also from volume to mass, so look up densities online but also based on data we have with commercial inputs).
  3. Write down assumptions. If you don’t have the info you need to do conversions in the survey need to make assumptions and need to write them all down.
  4. Take notes on each transformation of the raw data. Trust me, if you don’t take meticulous notes of how you transformed data and why it will be hard to summarize later (and even with notes its hard).  I think this has been the big difference between when I did my MSc and my PhD. I have made sure to take copious amounts of notes. I wouldn’t say it systematically reduces mistakes but it makes its 1000% easier to see where you made them and then be systematic about correcting assumptions if you need to as you go along.
  5. Look at all the data. Already, even though I am no where done finished processing there are things I have but didn’t anticipate. I am realizing there might be interesting information on # of inputs used, type of inputs, so more about management practices and then quantitative values.