Skip to main content
23 events
when toggle format what by license comment
Mar 23, 2021 at 19:00 answer added Guanglu Zhang timeline score: 2
Apr 13, 2017 at 12:44 history edited CommunityBot
replaced http://stats.stackexchange.com/ with https://stats.stackexchange.com/
Aug 24, 2015 at 14:31 answer added dave fournier timeline score: 7
Dec 15, 2014 at 2:40 answer added Adrian O'Connor timeline score: 3
S Jul 4, 2013 at 15:29 history suggested bjou
added 'least-squares' tag
Jul 4, 2013 at 15:27 review Suggested edits
S Jul 4, 2013 at 15:29
Jul 3, 2013 at 14:50 answer added Emil Friedman timeline score: 1
Jul 2, 2013 at 8:28 comment added Fixed Point @Glen_b Cool, gotta do a bit of reading (things mentioned in your post) and then some coding and then some numerical experiments to see how it turns out. I'll probably post other follow up questions later as separate questions. In the meantime, if you think of anything else for this question, let me know here. Thanks!
Jul 2, 2013 at 8:16 comment added Glen_b Please don't lose those additional questions, though - they deserve a post of their own.
Jul 2, 2013 at 8:16 comment added Glen_b I think my present answer shows that - in at least some circumstances - you can do better, and @whuber suggests something even simpler still than my process. I could go back and see how what I have performs on your data, but even as it stands now it does give some idea how to set such starting points up.
Jul 2, 2013 at 8:12 comment added Fixed Point @Glen_b Fair enough, I struck out the extra questions. So the questions is still, I have some data I want to fit using linear+gaussian form, can do I better than random initialization?
Jul 2, 2013 at 8:11 history edited Fixed Point CC BY-SA 3.0
added 51 characters in body
Jul 2, 2013 at 7:29 comment added Glen_b Well for starters, you're now asking about what you should fit, rather than specifying what to fit, as before. There are a number of other differences, some of which I'd think of as fairly significant. I'll look at changing my answer, but I think this one could stand as the original question and answer and your new parts where you are asking other things could be a new one. I'll leave that to your judgement for the present.
Jul 2, 2013 at 7:02 comment added Fixed Point @Glen_b Why is that so? Why does it look like a new question? As for the old question, we all whore for points ;-D and the old one has two upvotes, any way to merge it with this so I can keep those two votes too?
Jul 2, 2013 at 4:06 history edited Fixed Point CC BY-SA 3.0
deleted 3 characters in body
Jul 2, 2013 at 4:06 comment added Glen_b +1 for extra information, but now this looks like a new question. Incidentally, if you want to delete the previous one now I think that would be okay, it looks like you have now covered the additional information it had.
Jul 2, 2013 at 3:55 history edited Fixed Point CC BY-SA 3.0
added 8384 characters in body
Jul 1, 2013 at 7:55 vote accept Fixed Point
Jul 1, 2013 at 7:34 history edited Nick Cox CC BY-SA 3.0
deleted 10 characters in body
Jul 1, 2013 at 7:21 answer added Glen_b timeline score: 11
Jul 1, 2013 at 7:12 history edited Glen_b
edited tags
Jul 1, 2013 at 7:02 history migrated from mathoverflow.net (revisions)
Jul 1, 2013 at 3:18 history asked Fixed Point CC BY-SA 3.0