[00:24:44] Hey folks. Does anyone know if limn can handle normalized data yet? [00:25:11] e.g. I want a graph that plots a few different timeseries on top of each other based on a factor column. [00:25:26] Is that possible or does each set of values need to live in a column? [00:44:00] drdee: that paper was all over the web this morning [00:45:00] I like that someone is doing serious sociology of science to understand the use of p-values across large bodies of literature, but other than that - meh - it's the same old rant about blindly trusting a p-value that I've seen in different forms over several years [00:45:45] halfak: normalized data: no (AFAIK) you will need to preprocess the data [00:46:30] DarTar: You mean like the rant Zack and I give" [00:46:32] ? [00:46:32] booo [00:46:41] ha ha [00:47:19] read the paper and let me know if you disagree :p [00:47:29] I'm not sure how I'm supposed to build a dataset with an arbitrary number of columns. :\ [00:47:40] Dartar: can you send me the link please? [00:47:50] http://www.nature.com/polopoly_fs/1.14700!/menu/main/topColumns/topLeftColumn/pdf/506150a.pdf [00:47:53] I'm not fancy enough to have an irc relayer [00:47:54] sayhar ^ [00:49:07] Oh dude it's Ziliak [00:49:22] He co-wrote this: http://www.amazon.com/The-Cult-Statistical-Significance-Economics/dp/0472050079 [01:08:54] halfak: k I may have misunderstood the question then, do you mind following up by mail (copying milimetric who knows more about what Limn can or can't do)? [01:09:33] Sure. [01:09:40] DarTar: ^ [01:09:46] cool [01:18:00] Hey DarTar, sorry I haven't replied to the Trello thread. Vibha needed me to take of some other stuff first. [01:18:05] DarTar: Will try and do it today [14:12:21] hi [14:12:29] I was looking throuhg the operations puppet manifests [14:12:39] some ssl-related machines are mentioned there [14:12:54] I know this was a recent topic on the internal mailing list [14:13:56] in manifests/role/cache.pp [14:14:16] in this repo https://github.com/wikimedia/operations-puppet [14:21:01] and there's also a class called protoproxy::localssl in there which looks to me like something that would configure an nginx [14:27:32] (PS7) Nuria: [WIP] Changes tu support wikimetrics in vagrant. [analytics/wikimetrics] - https://gerrit.wikimedia.org/r/109676 [15:28:10] (PS8) Nuria: [WIP] Changes tu support wikimetrics in vagrant. [analytics/wikimetrics] - https://gerrit.wikimedia.org/r/109676 [15:48:05] (CR) Hashar: basic tox environnement (2 comments) [analytics/wikimetrics] - https://gerrit.wikimedia.org/r/110596 (owner: Hashar) [15:48:53] (PS2) Hashar: basic tox environnement [analytics/wikimetrics] - https://gerrit.wikimedia.org/r/110596 [15:49:06] (CR) Hashar: "Nuria, patchset 2 should work for you now." [analytics/wikimetrics] - https://gerrit.wikimedia.org/r/110596 (owner: Hashar) [16:04:21] (CR) Hashar: [C: -1] basic tox environnement (1 comment) [analytics/wikimetrics] - https://gerrit.wikimedia.org/r/110596 (owner: Hashar) [17:32:16] G'morning sayhar. [17:32:25] :) [17:32:26] morning! [17:32:33] I was just pulling up the email you sent me last night [17:32:51] :D I'm excited to see what you think. I hope I got the larger simulation right. [17:36:24] so I haven't gotten into your edits yet, but look what happened over the night: [17:36:46] http://cl.ly/image/3U1a431z441r [17:37:52] column 1 is aaron_simulation (slow version from yesterday), column 2 is aaron_boot, column 3 is a t test, and column 4 is a binomial confint (just for binary if donated - as a reference) [17:38:11] The 3 different types give pretty much the same results [17:39:48] Hmm.. Not not sure that I understand the meaning of the numbers. [17:40:09] But if you look at the test I ran, you'll see that it sets the CIs right for 95% match. [17:40:18] Also, I was able to run 1000 trials in about 5 minutes. [17:42:17] hmm [17:44:23] what are you doing with qbeta? [17:47:10] OIC [17:49:53] hot [17:55:03] sayhar. Sorry I missed your question. [17:55:17] I ran prop.test just in case. [17:55:52] Note that the 95% confidence is minutely smaller for qbeta than prop.test. I'm not sure if prop.test is using a normal approximation or if there some correction that it is doing. [17:56:20] hey halfak: did you see the note felipe posted to analytics-l? Pretty good timing [17:57:48] I did. I was considering if I should just send him the dataset or block off some time to complete a write-up. [17:57:53] DarTar: ^ [17:58:20] we should let the world know about two things (or 3) [17:58:39] 1) metrics standardization (which also includes "clean" registrations) [17:58:57] 2) your reconstructed dataset about user regs [17:59:28] 3) the possibility of sharing a view of EL data on account registrations on labs [17:59:50] I'm happy to do it, unless you get there first [18:00:09] moving to the conf room for the all staff [18:00:35] Sounds good to me. Just ping me when you get there so that we don't duplicate work. [18:00:53] No hangout! [18:01:01] Wharhgrhaaarble! [18:01:47] tnegrin, hengout plz [18:01:54] https://plus.google.com/hangouts/_/wikimedia.org/data-and [18:04:55] sayhar, ping. [18:27:02] hello [21:18:56] milimetric, any chance you can bump the public datasets directory for me from stat1 to stat1001? [21:19:37] i don't have any more rights on those boxes than you halfak [21:19:39] maybe less [21:19:52] i think the cron runs every 30 minutes [21:20:08] but i can show you what i mean without the specifics [21:20:21] Okay. Sounds good. Batcave or other? [21:21:30] milimetric, ^ [21:21:45] batcave, sure [21:21:53] http://goo.gl/1pm5JI [22:17:44] ok halfak: git pull in the limn directory [22:17:54] and you should be able to see the tsv url rendering as a graph [22:18:39] yay. ty [22:20:07] hi halfak how are you? [22:20:33] hey sumanah. not too bad. :) [22:20:48] halfak: any chance you're coming to PyCon in Montreal in April? [22:21:23] I wasn't planning on it, but that sounds interesting. [22:21:52] halfak: http://www.harihareswara.net/sumana/2014/01/19/0 lists some talks I find interesting-looking [22:23:06] * halfak reads [22:23:54] halfak: did it work for you as well? [22:23:58] just want to make sure you're set [22:24:40] Sorry for the delay. I'm not used to git. I usually work with mercurial. [22:25:34] no problem, but i really meant just "git pull" [22:25:53] You don't need to tell git how to apply the changes? [22:26:07] shouldn't have to do anything fancy in this case, no [22:26:14] Usually the patten in hg is "hg pull", "hg update" or "hg merge" if there's something interesting. [22:26:20] oh right [22:26:26] with git it would be: [22:26:29] git fetch [22:26:30] git merge [22:26:47] but in this case you don't need fine control over the merge since you haven't changed anything since the last sync [22:26:56] so the shortcut for git fetch --all && git merge is just "git pull" [22:28:02] You'd think that going back and forth between git and hg would be more straightforward. [22:28:27] Works! [22:28:32] ty milimetric [22:28:36] no CSV for me [22:28:41] cool [22:28:47] yeah, tsvs are better anyway :) [22:28:47] Ironholds: !!! [22:28:52] I used a TSV in a limn plot. [22:28:53] and it'll read json datafiles too [22:29:04] * halfak carries milimetric around on his shoulders [22:29:16] I'm much stronger on the internet than I am in real life. [22:29:22] hahaha, super happy to actually be of assisntance [22:29:26] as always - comments welcome [22:29:47] oh man, I just got called fat [22:29:48] :P [22:30:13] aww snap [22:30:57] have fun y'all, it's been a very long day, nite! [22:32:05] g'night milimetric [22:35:20] DarTar, got a sec for some limn advice? [22:35:51] talking to Ironholds, 20 mins? [22:35:55] sure [22:50:10] (PS3) Milimetric: [WIP] Run recurring reports using the scheduler [analytics/wikimetrics] - https://gerrit.wikimedia.org/r/112165 [23:07:09] halfak: I'm back [23:07:25] Hey! So I want to add a plot to the ee limn. [23:07:38] It covers many languages. New tab or is there a current tab you think makes sense? [23:11:58] halfak: you mean dashboard right? [23:12:04] dashboard -> tab -> graph [23:12:09] Oh yes. [23:12:33] ok, [23:12:33] so there's a secret multi language dashboard nobody knows about [23:12:41] http://ee-dashboard.wmflabs.org/dashboards/top10-metrics [23:13:11] probably this doesn't make sense if you're looking into non top-10 projects [23:13:13] and the name totally sucks anyway [23:13:24] but that's the closest we have to a cross-language dashboard [23:13:42] Vital Signs should hopefully replace all this [23:13:43] OK. I think I'm just going to add one for "growth" [23:24:07] DarTar, I think I just sent you a pull request on github for the new dashboard & graph [23:24:24] k checking [23:29:21] halfak: I merged and tested locally and it was not working but I guess it's because I haven't updated Limn in a while [23:29:46] Oh yeah. There was an update that milimetric did while we were testing. [23:30:02] do you know if the instance running on ee-dashboard is using the latest code?> [23:30:34] ah crap [23:30:53] latest version of limn breaks my local instance :-/ [23:31:24] woops [23:31:30] a new dependency on passport-mediawiki-oauth, apparently [23:31:35] I'm not sure what is running on ee-dashboard [23:31:42] "new" [23:31:54] I guess that is pretty new [23:31:56] I can try and merge the data files there and see what happens [23:32:08] hold on [23:35:03] halfak: http://ee-dashboard.wmflabs.org/dashboards/growth [23:35:20] is that how it should look like? [23:35:29] Adam Hyland loves you. [23:35:38] yay [23:35:58] ok, off to talk with sayhar, bbl [23:36:10] thanks again! [23:36:15] np