plasticbag.org

This is absolutely the last post on visualisations of plasticbag.org data unless someone sends in something else really cool – and this one is more of a clarification than anything else. Daniel Boyd created this really cool model of post time that had accurately predicted both the times I woke up and went to bed, but it had seemed to go a little pear-shaped as a visualisation at the time I switched to using MT. I’ll post the original graph below for those of you too lazy to look at it in its original context (slackers):

Anyway, the confusion emerged at the beginning of January 2003 when suddenly it looks like I was staying up all night and posting at completely random times of the day. I instinctively felt like this could not be right – and by instinctively I mean that I knew it wasn’t right because I’ve become mostly old and predictable in my early thirties. Enter Tom Carden again, who writes:

“The time of day chart is great, but I suspect that more careful analysis is needed here. For instance, I believe that the move to MT coincided with the enabling of comments. I suspect that’s why the time chart suddenly goes strange. Here’s one without the comments. It’s as you might expect – pretty much the same times, but less frequent.”

He’s also generated a visualisation that shows comments as green dots and trackbacks as blue. It’s harder to interpret these on images reduced so much, so if you’re gripped by this whole subject area you might want to check out his two larger-scaled graphs: Without Trackbacks and Comments and With Trackbacks and Comments. The smaller version looks like this:

By way of intepretation then, I think it’s clear that since moving to MT my posting has dropped enormously (there are other reasons for this of course – including my rather hardcore nine months working on Radio 3. Perhaps more interesting is that the same posting patterns during the day appear to have continued (first posts between eight and ten in the morning – last ones between 12 and 2am) but there is more deviation from this practice. Whether this just represents being at conferences in other time zones or something isn’t clear to me.

Looking at the comments and trackbacks, it seems clear that they’re much more consistent during the day and night – which probably reflects a fairly international audience (or a lot of insomniacs and filthy drunks). There does appear to be a bit of a concentration of comments at least between midday and six UK time – although that may just be a result of comments lagging after specific posts. That would also account for the fairly heavy striping in comments up and down the page – on occasion I seem to write something that a lot of people want to comment upon. The rest of the time – not so much… There are also some fairly clear stripes of total inactivity emerging – February/March-ish of this year seems a complete dead zone as does a good period around two-three months ago.

I think generally what this little project has taught me is that statistics and visualisations are are both really good fun and that they can expose to you patterns and behaviours and causalities that you may have suspected were there all along, but couldn’t be certain about. Along with projects like Audioscrobbler, I feel like I’m starting to get a strange and exotic new statistical understanding of my life. I can look at these diagrams and see myself in them. It makes me want graphs of the fibrous content of my excrement automatically generated by my loo – and a complete break down of the percentage of my floor which is covered with rubbish and books at any given moment in time. Somewhere in this stuff is self-knowledge!

Wow, so it’s nearly three in the morning, which is basically four in the morning since the clocks only changed a week ago, and I still appear to be up and awake and completely uninterested in sleep. I may as well take this opportunity to post a few more visualisations of the weblog data that I posted up a few days ago (Five Years of Weblog Data vs. The Visualisations). This batch pushes us in a few interesting new directions – some of which so interesting in fact that I have no concept whatsoever what they might signify or represent. In fact let’s start with those:

These two were undertaken by Dan Kaminski and were “run through the Phase Space Visualization process popularized by Michael Zalewski” using Phentropy. I will confess straightaway that I’ve spent a limited amount of time trying to work out what they mean, but remain as yet unsuccessful.

The second set of visualisations came in from Daniel Boyd and is a plot of when during each day I posted over the last five years:

The first three years here are astonishingly consistent and – as Daniel himself pointed out – the data would suggest that I go to bed between midnight and 2am and that I get up each morning around eight. This is stunningly accurate and highly representative of my normal behaviour. What’s less clear is what happens after the beginning of 2003 where the data starts to look remarkably less organised. I’m still not clear why this has happened – and in fact quite suspicious of how accurate that bit is. It seems to me more likely that some part of the data is corrupt around the beginning of 2004 than that my entire posting behaviour changed overnight. Except to say that yet again this phase shift seems to coincide with switching to using Movable Type.

Which brings me to the last visualisation. When I last posted these visualisations I pointed out that they seemed to suggest that the switch to Movable Type had resulted in longer posts and a drop in the rate of posting. In response Anil Dash mooted that Movable Type might be making my posts better thought out. Now of course, there’s a clear value judgement there, and one that is not necessarily correct. It’s quite conceivable that I’ve just got more wordy and long-winded and that MT has supported or even caused that shift. It’s not even clear that weblog posts should be better thought out – Matt has suggested to me in the past that another equally convincing model might be to think of MT as having broken the paradigm of incredibly fast and easy informal peer-to-peer publishing that Blogger created and was initially its USP.

So it’s just as well that we have Richard Soderberg out there. He decided to run a Flesch-Kinkaid reading analysis on each of my posts over the last five years and plot them on a graph. The solid darker line in the middle being the reading level in the US educational system that might be required to be able to read each post.

As you can see, the readability level of my site changes dramatically around the same time that I move to using MT. Over the first three years it seems to settle around eight on the Flesch-Kincaid scale. But after moving to MT, the average appears to be more around 10. More interesting still is that the extremes in both directions have started to diminish. There are few posts that plummet down to the infantile depths of the scale and few that stretch upwards towards unintelligibility. Richard’s also provided a useful CSV of some key metrics. When you investigate it, you’ll see that each post has its complexity described in as either childish, acceptable, ideal, difficult or unreadable. I wonder if the same metrics apply with writing for the web as for elsewhere (we’re normally told by usability people that sentences should be shorter and punchier and articles not as long when we write online) but if they do the site has definitely moved much more heavily towards the ‘ideal’ intelligibility rating and that does appear to coincide with a move to MT.

So there you go – three new sets of visualisations and a stunning revelation that they should put on the marketing bumph. That’s right, ladies and gentlemen, apparently MT actually makes you smarter!