Sorry for taking so long to get these results out. Of course, life had a hand in this, but I also ran into some problems in the process of separating usenet-based responders from web-based responders (read "the rest of the world"). As a result there will be a total of 6 different results files, as I explain below.
At first, I thought I managed to do a rough split of the data using question H1 ("how often do you at least READ rec.arts.anime newsgroups"), but I think some folks didn't realize that "rec.arts.anime news group" includes rec.arts.anime.misc, etc and so I had about 3% of folks who answered "never" subsequently saying they read rec.arts.anime.misc. Similarly, I had folks who said they read rec.arts.anime news groups subsequently answer they didn't read rec.arts.anime newsgroups in other questions, somewhere around 1-5% of them. That is, it isn't ideal. It gets worse.
Then I got a bit suspicious of the final usenet-based tally of 167. By the time the web announcement came out, a few weeks after the first raam announcement, I had about 75 returns. That means I got another 90ish after the web announcement, which is contrary to past experience: 1/2 - 2/3 of all returns are supposed to be due to the first announcement, and 2/3 to 3/4 from the 2nd. On the otherhand, I so botched up the announcment schedule this year that the web announcement may have effectively been the 2nd announcement. I also saw many more lurkers responded this year than in the past (H2). I hoped that was due to the web form not requiring e-mail. But there were a number of inconsistencies from previous years' results, sudden jumps which at first I thought may have been real.
Then I decided to look at just the 75 returns from before the web announcment, just to be sure. Sure enough those inconsistencies existed for concurrently, so they weren't real. So I cut the data in several different ways, and the conclusions always came out the same: there's a difference between 1) the population that make newsgroups home (but use the web), 2) the population that don't have much to do with newsgroups, and 3) a population that makes the web home (but uses newsgroups, which I'm calling "Mixed"). The latter threw me for a loop, until I saw a cluster of really obvious distinctions in the posting frequency. Aha, these are infrequent visitors to usenet, probably past regulars who've moved on carrying some of the characteristics of usenet readers and picking up some web reader characteristics. Then I had to recode my software (yet again!) to accomodate the necessary cuts (gone are the days of simple bean counting).
So, I've split up the subtallies into 3 files (one for each group, Usenet, Web, and "Mixed"), and done three comparisons in three more files (group 1 vs group 2, group 1 vs group 3, group 2 vs group 3). If I have the time I may tabulate those two groups of three in a pair of files for easier comparisons.
I have to think about how to get better separation next year; I'm going to have to be clearer in discriminating these forums, and maybe running two separate surveys is the only way to do this (but I have to work at minimizing repeat responses). It could be that by luck my delaying the web announcement was actually the way to go.
As noted in the results, only the quantities that are 2 or more error bars (standard deviations) are noteworthy, so I point only those out here. What we find is that there is a general similarity. The question is are the expected differences there?