[00:01:03] yeah, I'll get around to that someday... [15:00:32] Hi [15:00:35] I'm here [15:01:59] hey YuviPanda! [15:02:04] glad you could make it :) [15:02:24] Hey J-Mo [15:02:24] we're starting in about 2 minutes [15:02:25] Just IRC tho [15:02:29] Am outside [15:02:29] np. we'll be fine [15:02:43] Cool :) [15:03:10] J-Mo: tell people to just submit again if it is queued for more than 10s [15:35:22] quiddity: J-Mo do you have a link for the video? I just got back to a computer [15:35:58] https://plus.google.com/hangouts/_/hoaevent/AP36tYfEYDju0LyCqmDIQR45_cFK-lzPaSDLKepacIqJcKZN74JBAQ?authuser=0&eid=117126229120807344920&hl=en [16:15:53] J-Mo: more fun fact - we actually released search queries a few years ago, and then ran into the exact same problems and pulled it :) [16:22:05] am off to eat [16:22:06] brb [16:22:09] thanks! [16:23:50] whew. all done. Thanks again YuviPanda, for everything! [18:32:22] morning happy campers [18:32:42] hi - is there anyone who knows about design of experiments :-) [18:32:56] and also R [18:33:01] ;-) [18:33:41] OrenBochman1, I've got R, halfak knows a lot about experimental design (and a lot about R) [18:35:27] I'm trying to use the R package FrF2 (Fractional Factorial designs with 2-level factors) [18:35:49] ahh. never used it! [18:35:55] I tend to avoid factorising. It makes me sad. [18:35:58] and I keep getting errors [18:36:16] yep most people just do A/B tests [18:36:29] oh, I just meant the R concept of a factor [18:36:44] but there are many situations where it is inefficent or unsuitable [18:37:45] e.g. - with multiple factors/levels which also interact [18:38:19] uhhuh. Yeah; not my area, I'm afraid :( [18:38:22] hey leila :) [18:38:28] yo Ironholds. [18:38:33] :-) [18:38:41] perhaps leila knows this stuff [18:38:47] mmm, lemme read [18:39:01] (I may have to leave in the middle. will read it later if that happens.) [18:39:13] ok [18:40:11] I got to go soon as well [18:40:36] I noticed you may have backround in medical reseach [18:42:02] I'm interested in looking into designing a kind of experiment to test the effect of a combination of a number factors [18:42:59] e.g. wikipedia template messages: text, call to action strength, font size, color, icons [18:44:47] since each var can take a few levels. [18:46:03] the design needs to take a different level from each variable. [18:46:49] it is usually possible to avoid checking all the combinations [18:47:26] to find the optimal messages [18:48:37] OrenBochman1: I haven't worked with the function you referred to earlier. However, one way to handle this situation is to add th interactions manually? [18:48:57] so, if x_1 and x_2 are your variables, add x_1*x_2 to the model as well [18:50:14] also, what kind of error do you get, OrenBochman1? [18:50:53] what you suggest would happen as part of the factorial design [18:51:51] yup, I understand, what I meant is that you can use a simpler function, and add interactions manually, if this one is giving you a hard time [18:52:10] or let me know what your error is and I see if I can immediately get what the problem is [18:52:33] and the errors - I guess it depends on - if the intercation is linear etc and which factors are stornger [18:52:54] also some of these design require repettion for the weaker vars [18:53:11] I guess I'll have to research this some more [19:55:50] DarTar, Ironholds, a db question for you. revision.rev_len contains bytes CHANGED in a revision, not bytes added, yah? If so, is there a canonical approach to discriminating between a -2000-byte change and a +2000-byte change? [19:56:30] J-Mo, ah, yes [19:56:38] revision.rev_len is length in bytes post-change, not the delta [19:56:43] the delta is only store din recentchanges [19:57:05] so you can compute by comparing revision.rev_len to the revision.rev_len of revision.rev_parent_id [19:57:14] Ironholds: stat1002 now has /srv/aggregate-datasets and putting things there will make them public [19:57:42] YuviPanda, coolio! [19:57:49] now write things :P [19:58:00] Ironholds: and that comparison will yield the value of the change, positive or negative? [19:58:16] yup [19:58:24] although obviously it's a bloody pain to do programmatically [19:58:33] I recommend - and I can't believe I'm saying this [19:58:51] inner joining the revision table on to itself where rev_id = rev_parent_id [19:59:07] cool. do you (or milimetric) know whether Wikimetrics uses this approach when it delivers bytes added? [20:02:37] disregard ^^. answered. [21:03:54] hey kevinator_, do you see DarTar nearby? [21:04:58] no, he’s offsite today, working on strategy meetings with other directors and upper management [21:09:22] Gotcha. Thanks