[14:58:13] morning lzia! [14:59:24] morning Ironholds. how's it going? [14:59:33] whelp [14:59:37] I've been awake for 22 hours [14:59:41] only 4 more to go! [14:59:59] ha! making progress on the category analysis? [15:00:20] yes, I analysed that it was a dumb idea [15:00:29] there are two options: unplottable and unmanageable. [15:00:51] so I'm going to go with looking at disparities in article choices specifically [15:01:06] and in user behaviour around those articles (is there a long tail or a more even distribution?) [15:01:17] what do you mean by article choice? [15:01:44] in a random sample of 100,000 desktop edits, N willl cluster on the same article choices [15:02:05] in a random sample of 100,000 mobile edits, N/X will cluster on the same article choices [15:02:11] and the article choices are different in each group [15:02:28] are you focusing on one language/wiki? [15:02:35] in one group it's page 93432 that people cluster around, while in the other it's 3322. Massive simplification because I'm really tired but you get the idea [15:02:37] yeah, probably enwiki [15:02:54] the advantage of this is that it's independent of the circadian rhythms/geocoding work and so I can pull out a lot > 90 days. [15:02:54] have you started sampling? [15:03:06] not yet, I just got the circadian rhytmh plotting set up. [15:03:11] on which note can you send me that code? ;) [15:03:18] I'll hopefully tear through it at the hackathon tomorrow. [15:03:21] ...or whenever that is. [15:03:31] I /think/ uniform sampling won't give you enough results [15:03:47] define uniform sampling in this case? [15:03:59] that is, the same from both desktop and mobile? [15:04:06] that's why I'm going for >90 days. [15:04:10] plus, we actually have a control. [15:04:24] we can compare different random samples of desktop groups, which make up the vast majority, to see what the expected level of overlap would be. [15:04:43] so we solve for the problem of "okay, so all you've shown is that the desktop edits you picked THIS time don't line up with the mobile population" [15:05:39] I should shower. back in a tick! [15:05:59] suppose you have N possible candidates (editors) on desktop. are you thinking of uniformly at random choosing k of them, or 1) you want to have some weighting mechanism for choosing certain editors over the others or 2) you want to look at a specific subset of editors: those who have done 5+ edits in the previous month, etc.? [15:11:35] none of the above! [15:11:37] edits not editors. [15:11:57] the problem with picking editors is that there is no uniform distribution because there is so much more legacy the desktop editors have. [15:12:28] I would be saying "there are behavioural differences between these two groups, let's, for the purpose of sampling, assume these could not impact any of the metrics we're using to sample" :D [15:13:02] and now I am repeating your probable argument back to you ;p [15:37:22] CSCW proposal has a complete draft! [15:37:50] yay! [15:38:17] Ironholds, it's only a couple pages. Could I bother you to take a pass and comment? [15:39:20] I've not slept in...we're up to almost 23 hours now. [15:39:27] but if you're okay with "needs more unicorns" go ahead ;p [15:39:28] Oh. lol [15:40:04] PM'd link [15:40:15] ta [21:54:06] Hey Nettrom. It looks like user pages are down for UMN. E.g. http://www-users.cs.umn.edu/~halfak [21:54:13] Is it working for you from UMN? [22:01:21] lemme check [22:01:32] can't establish a connection [22:01:57] appears that I can ping the machine, though [22:02:17] * halfak is emailing operator [22:02:36] the main CS page works, though, not sure if that's also internal [22:04:10] email sent [22:07:26] thanks for getting that sent! [22:07:34] I have to head out, need to finish packing up my apt [22:08:14] o/ [22:08:33] o/