[fr] On ne peut pas prendre deux mesures au hasard, en faire un rapport, et espérer qu'il ait un sens. Un peu de rigueur scientifique, que diable!
10.02.2011: Seesmic recently took its video service down. I have the videos but need to put them back online. Thanks for your patience.
Video post prompted by Louis Gray’s Twitter Noise Ratio. I’m still somewhat handicapped and used up my typing quota this morning. corrections: measure time, measure distance (not “speed”) My graphs: 
Zoom in to the beginning of the graph: 
Attempt to spot trends: 
Not conclusive. See also: Stowe’s Twitterized Conversational Index — interestingly, Stowe became much more “chatty” on Twitter lately 😉 Update: The Problem With Metrics — a few thoughts on what metrics do to the way we behave with our tools. Confusing ends and means.
I seem to have said this a few times the last couple of days, and you touched on it briefly, the only really way to measure signal and noise involves assigning a value to tweets (something you touched on briefly in you post). Where I agree that it’s subjective, it’s the only way I see to determine a value in someone’s tweets (something I wrote about a few weeks ago http://seanreiser.com/node/265).
Hmm,,, It occurs to me at the very least, I think we’d need a simple “digg style” I like / I don’t like this tweet (not favorites… something that can be used a little less judiciously). With that we could build a simplified model based on trust matrix.
I really don’t think it’s a problem of statistics, it’s more a problem of not having enough data for the statistician to use.
Looks good! I like the video update and hope your typing capabilities come back strong soon!
I made the Twitter Noise ratio to show how I think about it, and clearly, others think about it in a different way. Why measure? I guess, because Twitter gives us the numbers. If they stopped showing # of followers, # of followed, etc., would we be okay with that?
And yes, it’s not perfect, but it certainly has people thinking about it. I hope new metrics debut.
The video was great! I left more comments on Friendfeed, but I think you’re dead on with your conclusions and analysis. What Louis provided wasn’t sufficient evidence, but it was a starting point.
Good stuff,
My main concern about this is:
– multiplying or dividing figures does not make a stat.
– all those “stats” are done on a small sample (the people I follow/who follow me) which is not representative
It reminds me when developers think they can do design or when designers think they can code.
I agree that a subject need not to be clearly defined to be discussed, however just spitting figures with disclaimers (à la Dave Winer) or arguing that since we have figures we should use them, does not bring much to the discussion.
It’s only noise.
I am not one to watch video posts normally but I am very glad I did in this case – excellent post Stephanie.
STephanie, nice post, oops video, usually don’t watch these.
I agree with you on the metrics problem and I have posted something about it here, not being aware of this post, of course my fault:
http://commetrics.com/?p=35
I feel metrics are great but the challenge is to find effective ones that address what you intend to measure – and yes, numbers by themselves mean little if anything.
And defining what is noise and what is not or value added versus noise is important and the time factor as we pointed out in the above post is an important control variable that must be measured as well.
Thank you for sharing this with us.