Back in April I posted data on three indicators of “influence” for ~400 science-focused Twitter users – based on David Bradley’s list of “Scientific Twitter Friends.” Intrigued to see how these Tweeps’ influence evolves over time, I will be updating these data periodically.
In this first update (aided and abetted by @ruthseeley – thanks Ruth!), the overall number of followers (both primary and secondary) of the SciTweep cohort has increased over the past two months – as would be expected given increasing interest in Twitter. There is the slightest hint of an increase in overall Social Capital. But this is marginal, suggesting that SciTweeps are not deviating substantially from Twitter-wide trends in increasing followers.
These data are available on Many Eyes to play around with (see the screencast below for tips on how to mess around with the bubble chart). You can even download the original data here and dive deeper into it…
The dataset is reasonably large and no doubt holds a multitude of insights for those dedicated to mining it (although with only two date points, it is still lacking in depth). But rather than go into deep analysis here, I thought I would restrain myself and simply focus on the distribution of the three indicators amongst the group, and changes over the past two months. (Details on the three indicators of “influence” are covered in the April SciTweet blog).
In looking at these figures, please bear in mind that the group of ~400 Tweeps is one of convenience – it does not represent all current science Tweeps, and indeed overlooks some key figures in the Twitter community. But for the purpose of following a group of science-focused Twitter users over time, it serves its purpose well.
Looking at SciTweep followers, there has been a shift in the modal number of followers from 133 in April to 422 in June – although these figures are approximate given the step-size used. Clearly though, there’s been an increase in the number of people following most of our SciTweets.
There has been an overall increase in secondary followers over the past two months, although not sufficient to lead to a definite increase in modal value in the plot above.
The social capital distribution has barely shifted over the past two months. As this is based on the ratio of a SciTweep’s secondary to primary followers, it provides a measure of how the SciTweeps are faring compared to the rest of Twitterdom. A significant shift to the left or right would suggest the cohort of science Tweets loosing or gaining influence compared to other Tweeps. Given the similarity between the two distributions above though, it seems that the SciTweeps are holding their own, but not showing appreciably different changes in influence compared to other tweeps.
Finally, I thought it worth posting a quick screencast of how to navigate round the bubble charts on Many Eyes. Enjoy:
[flashvideo file=/movies/20090623/SciTweet_Influence_June09.flv image=movies/20090623/SciTweet_Influence_June09.jpg width=590 height=682 /]
Update 6/23/09: Bubble Charts updated with correct data for @maverickny (formerly – and erroneously – listed as @maverick_NY)
The data shown here are derived using Twinfluence.
Where the number of second order followers topped out on Twinfluence, it was capped at 30,000,000
My thanks to David Bradley for compiling the list of “Scientific Twitter Friends” in the first place. This is largely a self-selected list of science-types on Twitter, and in no way represents the full scientific community there. But it does provide a highly useful cohort of people who profess to have a science-perspective, and can be tracked over time. This series of analyses uses the list as it stood mid-April.
A quick word on the plots: These are a rather crude way of presenting the data, but provide a good qualitative indication of distributions and trends. The number of science Tweeps represented by each step in the plots represents the Tweeps with primary followers, secondart followers or social capital lying within the range of the step. As the horizontal axis uses a logarithmic scale, the range of values covered by the steps increases dramatically going from left to right. As the data are roughly logarithmically distributed, this makes visualizing, comparing and analyzing the numbers easier. But care should be taken when interpreting the plots, given the logarithmically compressed horizontal axis. In particular, modal values are dependent in part on the use of a log-axis, and would be different if the data were plotted with a linear horizontal axis.
And finally, many thanks to @ruthseeley for help in running SciTweeps through Twinfluence – a finger-numbing task!