Sunday, January 17, 2010

On the Anastomosation of Comment Threads

Presumably you know what I'm talking about already; if not this is unlikely to be of interest.

Teh Cephalopodian Overlord of Pharyngula conducted a fiendish and cruel experiment on January 8, 2010: two active comment threads were brute-force/microsurgically combined, an unnatural procedure jargonously known as anastomosis.

Quoth Teh CO at the time:
So there I am, holding two tubes on the internet that are pulsing and spewing, ready to cut both off…and what is any scientist's natural inclination in such circumstances? Why, to take the severed ends and suture them together and see what happens! That is this new thread, an anastomosis between two, count 'em, two old threads. I wonder…will it explode?
and, in alt-text:
Multiple threads fused into one gigantic monstrosity! Chaos reigns!

We now have sufficient data for a preliminary investigation of the early results from Dr. Myers' (if that is his real name) malicious "experiment".

Background: It was clear very soon that an actual explosion was not forthcoming, falsifying the only explicit hypothesis offered by Myers (itihrn) (though only as a transparent and pathetic post-hoc attempt at rationalizing his sadistic procedure with a thin veneer of some slimy synthetic material that smelled vaguely like scientific method). In the comments, somebody posting as "Myers" offered another hypothesis, albeit one positively dripping with sneering, callous, and--let's admit it--irresponsible disregard for the record-setting, award-winning Thread Everlasting and its threadizens:
More like "squish...SPURT...spurt...spurt.........spurt..................spurt.............sssssss"
Perhaps the patient will die on the operating table.
Well, no, Professor (if that is your real profession). Hell no.
The Thread, as you well know, lives on.
[But in full disclosure, my initial predictions, as well as a few others, were also somewhat exaggerated.]

More rational analysis was soon forthcoming, however, and I summarized my initial qualitative findings here.

What, however, of quantitative effects?
Although I was initially skeptical, the graphed data raised again, as from the dead, like Lazarus or Jesus or a zombie or vampire or whatever, the question: Is it possible that commenting rates were somehow affected by the Merge even if The Thread choked out its weird appendage topically? This is the question addressed hence.

Here are the relevant available data, abstracted from the Thread-history database, in temporal and threadual context:

Fig. 1: Detail of Thread Everlasting history documenting the context and aftermath of anastomosation. (click to enlarge)

These data, chosen to provide similar sample sizes for the pre-anastomosis threads and all available data post anastomosis, were first compared with simple linear regression:

Fig. 2: Linear regression analysis of data shown in Fig. 1. (click to enlarge)

An eyeballed Analysis of Covariance (no, I am not cranking SYSTAT up for this) suggests that the commenting rate (slope of the line) post-anastomosis (158 comments/d) was intermediate between the two pre-anastomosis rates (82 and 206 c/d), and about double the pre-anastomosis rate of The Thread itself, suggesting that anastomosation did indeed have a quantitative effect on commenting rate in this case.

However, it could be argued that commenting rates fluctuate naturally over time (viz. the frontpage effect at the beginning of each subThread), and that (as a result of such fluctuation) the Wieland-appendage data were poorly represented by a single line. I therefore ran a second, more restricted analysis on commenting rates immediately before and after the Merge:

Fig. 3: Proximity-restricted linear regression analysis. (click to enlarge)

This analysis only exacerbates the results of the inclusive analysis. Post-anastomosis commenting (167 c/d) was intermediate to the rates of the two threads pre-anastomosis (66 and 253 c/d), and the similarity of the slopes for the inclusive and restricted post-anastomosis lines shows that this effect was persistent in time.

Conclusion: This first, and (dog willing) last, experiment in comment-thread anastomosation suggests that the procedure has a lasting temporal effect on commenting rates, but probably not in topical subject matter. More research is desperately required (and at the same time ethically repulsive) to determine whether the observed trends are case-specific or generalizable.

No comments: