October 2009


So, to start with the bad news.  Neither James or I were particularly surprised to hear this week that the ESRC turned down our application for Rescue Geography 2, working on the live redevelopment process in Kidderminster (see the ReWyre site for details of the work the council are up to).  Two of the five referees loved it, two thought it was good with a couple of issues which needed resolving and one refereee absolutely hated it, though their objections seemed pretty irrelevant/missing the point.  Ho hum, this is the way things go in academia, particularly with grant funding being slashed at the moment.  It’s a real pity, because we’d have really liked to work with Ken and his team on what’s going to be a really exciting regeneration project.  So it goes.

On the cycling side of things, everything seems to be progressing smoothly.  I hooked up with Dan last week and we agreed the details of the exhibition we’re going to do.  It’s kinda exciting to feel like you’re commissioning art – perhaps I ought to get myself some kind of salon and have gatherings of the bright and the beautiful for the drinking of absinthe and exchange of ideas.  We agreed that this should take place in May, partly to give us time to organise everything and partly so that the weather’s a bit better, which is pretty crucial for something which is primarily going to be held outdoors.  So I’ve given Dan the contact details of all the participants who ticked the box saying they’d be happy to take part in a photography project and I guess we’ll see what happens.

I’m giving this talk on Monday, just an in-house seminar thing, but it’s forced me to really have a good look through the data and see what’s what.  The main thing has been that it’s really forced me to get to grips with some of the higher functions in NVivo, such as being able to analyse stuff by participant characteristics (e.g. gender).  As a result I’ve managed to produced an awful lot of graphs (after an awful lot of swearing trying to work out how to tell NVivo what I wanted it to do).  Just a quick sample of the kind of thing:

A graph showing themes discussed against the gender of the speaker

To produce this I rescaled the actual number of times a theme was mentioned to correct for the fact that there wasn’t a 50-50 split of men and women in the sample, so the numbers on the left hand side are indicative rather than real – otherwise the fact that there were 17 men and 11 women in the sample would have made it harder to identify themes were women commented particularly strongly.  For this graph I’ve picked out specific themes that showed a significant skew towards women.  You can then drill back down into the quotes to see what’s going on, say for example on cycle infrastructure, participant 10 talking about how difficult it is for her to lift the bike up the stairs off the canal at University Station.

I’ve also produced this map

An example of how GPS can be used in a slightly dodgy way

which gives a good example of the ethical issues raised when using GPS with participants.  Clearly I can identify people’s houses from the GPS track (although I exclude these from the maps on the website).  This means that I can map participants homes against, in this example, the Index of Multiple Deprivation and use this as a proxy for social class.  Now at no point did any of the participants sign up for the GPS data to be used in this way so it would be incredibly unethical to take participant’s home’s IMD score and use this as one of the modes of analysis within NVivo.  This particular example isn’t so problematic on this project as most participants live close (i.e. cycling distance) to the Uni and most of the areas around the Uni are pretty uniformally non-deprived, but on a different kind of project it could be a pretty dodgy way to make certain kinds of assumptions about participants.  Obviously with the scale used on this map and the anonymity it’s just illustrating a point rather than bending any ethical rules, but these are the kinds of things you have to think about when engaging in these kinds of technological projects.

Advertisements

Okay, so term has started now and it seemed an appropriate place to stop data collection, with some 28 people having kindly taken part – thanks to all. Since then I’ve been busy starting work on the analysis. On the Digbeth project I left a lot of this to Cosmic, but this time round I’ve been the one who has grappled with NVivo. For those who don’t know, NVivo is a piece of software which allows you to identify patterns, topics of conversation and themes within interview transcripts. Well, actually it does a whole lot more than this, but that’s what I’ve been doing with it.

I’ve not used NVivo since I took a training course back in, heck, early 2001. Back then I kinda saw the point, but didn’t really have an application for it. I got the latest version onto my shiny new computer and put all the interview transcripts in to start work. Almost instantly I was a complete convert – it’s an amazing piece of kit. Unfortunately it does require that you spend a lot of time sitting working through your transcripts. This is mostly pretty tedious – you start to decide a bunch of different categories, highlight text and ‘code’ it into that category. Categories might be, say ‘traffic’ or ‘senses’. Then you refine these, breaking them down into subcategories. So for ‘traffic’ I talked about things like pedestrians, other cyclists, danger and so on. By the time you’ve gone through all your transcripts you’ve got a bunch of different themes identified and you go back through making sure that the first few you did are coded against the themes you’ve identified by the end.

Like I said, pretty tedious.

What you get at the end, however, is a really interesting breakdown of what people were talking about and the common themes that keep emerging. This is particularly interesting on a project like this one where I didn’t give the participants any guidance about what to talk about aside from “what’s going through your head as you ride”. It’s a good way of starting to find out what kinds of things are really important when cycling. Or when commuter cycling at least. You can also analyse these things against the characteristics of your participants. This allows you to answer questions such as whether women and older people talk about ‘danger’ more than young men.

I haven’t done this bit yet and it may not turn out to show anything particularly interesting. One of the things about having this kind of software is that you do tend to play around a bit, looking at a whole bunch of different things because you can do quickly what once might have taken a couple of days so you might not have bothered. I have, however, finished the initial coding, a mere week and a half after I started on it (amazing how much you can get done if you close your door and turn your email off). Here’s a snapshot of the particularly interesting themes which came out strongly from the 28 participants:

Theme Number of participants commenting Number of comments
Comparison to other transport modes 17 27
Cycle infrastructure 24 116
Road surface 20 59
Animals 17 56
Weather 28 98
Hills 25 68
Accidents 7 10
Speed 22 48
Pleasure 23 49
Smells 9 17
Traffic danger 21 61
Junctions 24 76
Other cyclists 23 45
Pedestrians 26 81

Now there’s no getting around the fact that reducing complex, dense participant narratives down to a series of themes runs the risk of oversimplifying and concealing major parts of the story. Clearly one has to return to the quotes to get at the real depth of the material, but it’s useful to be able to zoom out – to use a mapping metaphor – and get a sense of the broader patterns at work.

On the subject of maps, I’ve reworked the cycling maps page on the site. All the transcripts have moved to a separate page and the top maps page is playing host to the analysis. So far there’s only three things there to look at: places where people rang their bike bells; places where people talked about different animals; and comments people made about pavements (mostly about riding on them). I particularly like the animals map downloaded into Google Earth, seeing the names of different animals floating across the landscape. As always with the ‘public geographies’ approach I’m posting up the analysis as I do it.

Got to get a shimmy on with the analysis too, as I’m presenting preliminary findings at a departmental lunchtime seminar on 26 October. This seems like a ways away, but you’d be amazed how many utterly pointless meetings I have to sit in between now and then which get in the way of analysing/thinking/writing. Ah, the joys of a real job…