How Very Meta

Coding Data September 19, 2008

I am slowly slogging through the coding of my survey data.  It is TEDIOUS in the extreme (though… let me plug the Coding Analysis Toolkit from QDAP, which is a lovely web-based tool for coding data) but it is still a very time-consuming process (at best, 4 hours to code about 900 items).

One of the hardest things for me has been coming up with good codes.  I mean, I know what I WANTED to get at with each question, but translating that into a reasonable set of codes (even allowing myself to choose multiple codes per unit) was sometimes an interesting process.

The question that has presented the greatest challenge has actually been the one in which I asked WHY people had categorized the blogosphere as a whole as public/private/both/neither.  I THOUGHT I knew what I wanted to get out of that question and sat down and developed a set of 10 codes based on a combination of my conception of the question and a relatively shallow perusal of the respondent’s answers.  When I started actually coding the data, though, I found myself shaking my head because the codes simply weren’t working… they weren’t lining up with what the respondents had said, and they certainly weren’t lining up with what I thought I wanted to get out of the question.

So I stopped, and I backed up.  I sat here in my office with my eyes closed and really asked myself what I wanted to get out of that question.  And I realized that I was trying to get at a LOT of things.  Within that question, I’m really asking about broader issues of public and private (is the blogosphere by definition public?  how does searchability play into that?  etc.), about blogging practices (do bloggers use friends-lock or other technological means of controlling access, do they deliberately avoid certain topics, do they obfuscate their offline identity?), and about the goals of blogging (to communicate with others, to convey information, to gain an audience).

So, wow.  That’s a lot of stuff from what I initially thought was a really simple question.  Now, though, I think I have a better set of codes, ones that are much more in line both with what the respondents said and with what I want from those responses.

I think the biggest thing I’ve learned from the data analysis process so far is just that.  Be willing to back up and start over if something isn’t making sense.  (Funny that I say the same thing to people who are impressed by my skills as a seamstress – “you just have to be willing to rip it out and start over if it isn’t working right.”  Why hadn’t I thought to apply that advice to my dissertation?)