Wednesday 23 April 2008

Coding data the tagcloud way

Health warning for research methods purists: THERE IS VERY LITTLE SCIENCE IN THIS POST (though there is a healthy dose of pragmatic common sense)

Since we introduced a personalised online space for students 3 years ago, we have been undertaking an annual survey to ensure that the student voice is central in informing future developments. With approximately 2000 responses each year and a mix of 20 quantitative and qualitative questions the exercise is highly valued. I have to admit that I personally look forward to getting my hands on the data and getting a feel for what students value. I am also a closet (or not so closet some might say) data freak, but it is a double edged sword - sleep is often sacrificed for the greater good of coding the data and it can take over chunks of my life.

Sooooo, on the bus on the way home to get stuck into coding more data, a strange thought crossed my mind. If I input the responses to a tagcloud generator would it do some useful coding for me? Obviously this is not the same a reading the full comment, and full qualitative quotes are the most useful in how we disseminate that data, on the other hand, could it give me a quick overview of the response to a particular question. So I resolved nothing ventured nothing gained.
I decided to use the free-text service in tweetclouds (mostly cos tweetclouds, for me, is aestheically pleasing and it lists them in alphabetical order - useful for next bits...ie performing search on the words in Excel that catch my interest to check context)

I chose my favourite question on the survey that asked:
How would you describe shuspace to a prospective student?

and this is what I got:




(click on image to see larger size)

But look! look! - cloud created in 20.6341 seconds (my coding exercise will take considerably longer) and I now know that "information" occurs 421 times, "useful" and "easy" feature highly (though without checking, of course, I can't testify to whether they are prefaced with "not very" but a quick check into the data can tell me that) Similies are also easier to identify as are things that might seem surprising.

In the interests of science I will code the question and see what that tells me, but if I compare and, as I suspect, the trends are similar then the 20 second coding clouds may well entice me further ;-)

2 comments:

gs said...

ha! that's just brilliant. though suspect that it will disturb your sleep patterns, rather than return valuable sleeping time to you.

[are we playing student-voice-innovation poker, btw? ie, "i see your twitter research, and raise you tweetcloud analysis"? 'cos if we are playing...i so fold!]

Louise said...

thanks liz - i thought you'd like it - not playing all that other stuff poker cos i'd definitely lose. on the other hand enjoying the challenge of being made to think a bit more laterally and you know how i love the little clouds ;-)