Wednesday, January 17, 2007

Data hacking Part n

This never seems to end... It really is funny how every batch of new data is just different enough from all the other datasets so that it needs to be done either by hand or by rewriting some of the tools I use for data processing.

I've just spent a fun morning in trying to figure out why some files always get corrupted when emailed, and while I still have no clue as to why, I've at least found a way to prevent it. Hopefully the data processing goes smoothly once the uncorrupted files arrive; this is one of those projects where I'm just a data-monkey and wouldn't want to spend too much time with it.

I was going to write about another experiment earlier, as I was pretty excited about it, mostly because it failed, completely and spectacularly. And yet it's probably the most important batch of data I have collected.

We were doing an experiment in a day centre for autistic people, and were mostly interested in looking at some aspects of the timing and synchronicity of their interpersonal interactions. We know that autistic people have difficulties in interacting with other people, although there are several theories as to what exactly is going wrong. Also, as autism is not just a single condition, but a wide spectrum of disorders and as comorbidity with other developmental disorders and learning difficulties is high, no two cases are the same. No two people are the same, of course, but what I mean here is that "autism", while a useful label, affects people differently and therefore it is problematic to do experiments comparing for instance autistic and non-autistic populations.

There is a more defined subgroup in the autism disorder spectrum, namely those with Asperger's syndrome. People with Asperger's (like probably Rainman) are usually otherwise "highly functioning", often very intelligent, and for these reasons most autism research using statistic inference and experimental designs are done with people with Asperger's. In our study, this wasn't the case...

So, our idea was to get the experiment mobile (i.e. use laptops, portable speakers, battery-power the instruments etc.), get a train to this centre and get as many people as possible to try out our experiment. They were having music and dance sessions anyway, so our drumming-based experiment fit in their schedule and routine well (which is very important). My partner in crime knows all the people in the centre, as she volunteers there outside the term, and so everything went smoothly.

And then the experiment started. We had a plan for the kind of data we needed; we were going to compare two factors, and otherwise keep things very simple (experiment design -wise). We had also thought out a strategy for giving instructions, doing practice rounds etc. so that our participants would understand what we wanted them to do and would be comfortable doing it.

It became pretty clear that some of them couldn't do the experiment at all, and others probably could but for some reason didn't want to. The only person who could and would, needed to leave before completing the experiment, as his transport arrived... But rather than sinking into despair, we just decided to change plans completely. Which is not usually a good idea in a small project like this that is part of coursework. But, we didn't really have a choice. All our participants were interacting musically, or had a specific take on musical interaction. Each of them would make an interesting case study, and each were different (although a couple of groups could be formed, for instance the "would but couldn't" and the "could but wouldn't".

Nevertheless, I'm going through the data we collected during that day, just to see if there's anything there that could illustrate the descriptive case reports. This is partly because it is interesting, partly because I'd need to find some quantitative data to be dealt with, as the course requirements are like that.

I know, this is a distraction from my work, and I don't really need any new ones, but it's also stuff I need to do and that interests me, so I'm spending one more afternoon with it.

No comments: