I have blogged twice, here and here, about the iSentia election index. There were a number of questions around the accuracy of the numbers they were including. After contacting iSentia I didn’t really have much of a clearer view of what was being counted and what wasn’t. But what I could gather was the it included mentions of either their name or handles on Twitter, and possibly mentions of their names on Facebook, but that wasn’t clear. So just like with last weeks figures, I have used Topsy to compare the figures iSentia provide against a different source.
Firstly iSentia’s figures for this week:
In reverse order:
Hone Harawira. iSentia score 495. Topsy score 289.
Jamie Whyte. iSentia score 902. Topsy score 957.
Colin Craig. iSentia score 1526. Topsy score 1529.
Metiria Turei. iSentia score 2688. Topsy score 2290.
Winston Peters. iSentia score 2945. Topsy score 1671.
Russel Norman. iSentia score 3029. Topsy score 2726.
Laila Harre. iSentia score 3150. Topsy score 2405.
David Cunliffe. iSentia score 8202. Topsy score 7413.
John Key. iSentia score 18390. Topsy score 12958.
Once again it looks like iSentia’s stats are all over the place. The ratio between David Cunliffe’s iSentia firugre and John Key’s Topsy figure looks about right. This is based on it including mentions of John Key, as well as @JohnKeyPM. I know from my feed that a lot of people on the left talk about John Key in plain text, as opposed to mentioning him. I think there are a few more mentions of David out there, where he is refered to as DC (to save space), or there are mistypes of his last name. I think iSentia’s figures for John Key are way over blown. On the other hand I think there might be a few more mentions of John KeyS that are not included (quick Topsy search including John Keys brings up a result of 13,446. Which is about 500 more than before). iSentia seem to be over reporting some of the smaller party leaders though. This week it looks like Winston is the main one who has been overreported. The rest that have been all look to be over reported by amounts that would fit with the slightly different time frames being used by Topsy and iSentia.
Overall iSentia look to have upped their accuracy from last week. Metiria and Laila both look to be in their rightful places. But the figures for John Key still look to be wildly out. It would be great if iSentia could provide a public explanation of how they are collecting the data so its usefulness can be judged.