Timeline for Which test is better suited to compare averaged versus single data sets
Current License: CC BY-SA 4.0
53 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| S Jan 21, 2021 at 19:33 | history | suggested | Benyamin Jafari | CC BY-SA 4.0 |
grammar fixed.
|
| Jan 21, 2021 at 17:02 | review | Suggested edits | |||
| S Jan 21, 2021 at 19:33 | |||||
| Jan 21, 2021 at 16:28 | comment | added | kjetil b halvorsen♦ | The additional image urls are no longer working ... | |
| Jan 21, 2021 at 16:28 | history | edited | kjetil b halvorsen♦ | CC BY-SA 4.0 |
edited body; edited tags
|
| S May 5, 2014 at 22:38 | history | bounty ended | Glen_b | ||
| S May 5, 2014 at 22:38 | history | notice removed | Glen_b | ||
| May 4, 2014 at 3:25 | answer | added | Alecos Papadopoulos | timeline score: 2 | |
| Apr 29, 2014 at 23:03 | answer | added | Amit Moscovich | timeline score: 2 | |
| Apr 28, 2014 at 8:37 | comment | added | Py-ser | Ok, never heard about that, I will try to read something about. And many thanks for opening a new bounty! | |
| Apr 28, 2014 at 8:21 | comment | added | Glen_b | On the original question, I now think a functional data analysis (fda) approach might possibly be a useful way to look at the question, but I am not the person for that. | |
| S Apr 28, 2014 at 8:20 | history | bounty started | Glen_b | ||
| S Apr 28, 2014 at 8:20 | history | notice added | Glen_b | Canonical answer required | |
| Apr 28, 2014 at 8:15 | comment | added | Glen_b | Not as an approach to the original question, just on the question "is the appearance of coincident lines in that plot consistent with chance, or does it indicate clustering"... (if you were of a mind to pursue that line of thought.) | |
| Apr 28, 2014 at 8:01 | comment | added | Py-ser | Thank you. So, your suggestion is to go through Monte Carlo simulations? | |
| Apr 25, 2014 at 5:49 | comment | added | Glen_b | Thanks, that says a bunch. First, not much evidence of a few clear clusters, but there are a few places where there's a number of lines that are close to coincident. That may or may not be just chance - we might need to simulate to see the extent to which that would occur if they were all from one (noisy) population. | |
| Apr 25, 2014 at 4:37 | comment | added | Py-ser | Ok, I edited the original topic to add a new plot. It is not the best, but I think it gives the idea. If this is not the case, please tell me. | |
| Apr 25, 2014 at 4:36 | history | edited | Py-ser | CC BY-SA 3.0 |
added 97 characters in body
|
| Apr 25, 2014 at 1:37 | comment | added | Glen_b |
For my example plot (done in R), I set the color as col=rgb(170,170,170,30,maxColorValue=255) which is a medium-light grey set fairly transparent. If you find any single line segments off by themselves, they're quite faint, so it takes several lines to be (nearly) coincident to show up clearly.
|
|
| Apr 25, 2014 at 1:11 | comment | added | Py-ser | Oh I see. Different levels of the red line is due to different normalizations. The one used in the single profiles is the "right" one. | |
| Apr 25, 2014 at 1:05 | comment | added | Glen_b | Something like this is what I had in mind, so that darker grey areas are more suggestive of clusters of profiles. | |
| Apr 25, 2014 at 0:31 | comment | added | Glen_b | I will mock up an example. But why is the red line at a different height (y value) in the individual profiles than the first image? | |
| Apr 24, 2014 at 2:09 | comment | added | Py-ser | I mean: how can I show all of them in just one plot? There are hundreds of these profiles, If I show them all together you will be not able to recognize anything, any behavior, just some messy points. | |
| Apr 23, 2014 at 14:09 | comment | added | Glen_b | I'm afraid I don't follow your comment at all. | |
| Apr 23, 2014 at 10:29 | comment | added | Py-ser | All of them is not convenient: with such a binning, necessary to understand the profile's features, it would be just a mess. | |
| Apr 23, 2014 at 6:52 | comment | added | Glen_b | I mean "show them all on one plot". I may be able to mock up something like what I mean, but I can't do it at the moment. | |
| Apr 23, 2014 at 4:44 | comment | added | Py-ser | Sorry I missed the previous comment, but I don't understand your suggestions. The folded profile is the average, yes. What do you mean by "showing the individual ones"? They are showed in the linked figures. Also, what do you mean "clusters of individual profiles"? | |
| Apr 23, 2014 at 4:34 | comment | added | Glen_b | My earlier comment suggests one way to provide more detail - but for that I'd suggest subtracting the red line and perhaps using transparency so that clusters of individual profiles might be more obvious. | |
| Apr 23, 2014 at 4:31 | comment | added | Py-ser | @Glen_b, thanks. What kind of details could be more useful? Also, the figures are very explicative I hope. | |
| Apr 23, 2014 at 3:50 | comment | added | Glen_b | It might pay to add more detail and give this a day or two before I try to promote it again. I don't know how I missed the period expiring. | |
| Apr 23, 2014 at 3:46 | comment | added | Glen_b | The folded profile is an average of many? It might be worth showing the individual ones in a lighter color (perhaps grey), joined by lines, then what you have there over the top. | |
| S Apr 23, 2014 at 3:16 | history | bounty ended | CommunityBot | ||
| S Apr 23, 2014 at 3:16 | history | notice removed | CommunityBot | ||
| Apr 22, 2014 at 1:37 | history | edited | Py-ser | CC BY-SA 3.0 |
added 265 characters in body
|
| Apr 22, 2014 at 1:23 | history | edited | Py-ser | CC BY-SA 3.0 |
Added picture
|
| Apr 22, 2014 at 1:09 | answer | added | Hao Ye | timeline score: 2 | |
| Apr 18, 2014 at 2:54 | comment | added | Py-ser | @Glen_b, still nothing. Do you think the question is badly posed/not interesting/improvable ? | |
| Apr 15, 2014 at 3:36 | comment | added | Glen_b | Details here. More info here. Bounties are useful, but you may want to wait until you have a little more reputation before spending it on one. In my case, 50 reputation is no big deal, I can earn it back by answering a couple of questions. If you look at the featured tab on the front page, you'll see your question is on there (near the end); it will climb as the week goes on. I have 2 bonuses running at present. | |
| Apr 15, 2014 at 3:34 | comment | added | Glen_b | Py-ser Once you pass 75 reputation (I think) then after any question has been up 48 hours, a link appears under the comments below the question, but before the first answer (if any), which says 'start a bounty' and has a sort of pinkish background on my monitor. If you click it, you get some choices about how much bounty and which category of bounty to set, and you can add some text if you want. You can play with it up until the point it says something like "are you sure? This cannot be undone", so you can safely go to the "why are you awarding this bounty?" screen, say, and then back out. | |
| Apr 15, 2014 at 2:20 | comment | added | Py-ser | Thank you. I never promoted a question, I don't really know how it works. Let's see what happens and if it is worthy. Adding more details is good: I myself learn a lot by that. Thanks again. | |
| Apr 15, 2014 at 2:16 | history | tweeted | twitter.com/#!/StackStats/status/455892428856492032 | ||
| Apr 15, 2014 at 1:56 | comment | added | Glen_b | Actually I wish I'd done it two hours ago - I had reputation going down the drain because I hit the 200 point limit yesterday. A bonus would have been the perfect thing to spend the extra on without costing myself anything. | |
| S Apr 15, 2014 at 1:53 | history | bounty started | Glen_b | ||
| S Apr 15, 2014 at 1:53 | history | notice added | Glen_b | Draw attention | |
| Apr 15, 2014 at 1:53 | comment | added | Glen_b | Happily (and done). You may encounter some need to further clarify as this goes on. It now has a bonus, which hopefully will get it some attention over the next week. Depending on how things go there's always the possibility of doing it a second time. | |
| Apr 15, 2014 at 1:24 | comment | added | Py-ser | @Glen_b, if it is still your will, please let's promote this case :) | |
| Apr 13, 2014 at 2:18 | comment | added | Glen_b | Okay, this is a very interesting problem, but one probably at the margins of my expertise. If you have no good answers within 48 hours of the original post (about 30 hours from now), give me a nudge (@Glen_b me), here or on chat. I may be able to promote the question. | |
| Apr 13, 2014 at 2:14 | comment | added | Py-ser | Thanks, now it is clearer to me. Looking by eye, it seems to me that there is at least one "class" of profiles that is different from the average one. By "class" I mean that they look like similar each other (they have similar patterns), but still different from the average. However, this is just by eye, so I want to check if they are significantly different from the averaged profile. | |
| Apr 12, 2014 at 13:46 | comment | added | Glen_b | I'd have thought that the use of the word 'each' conveyed that I didn't mean 'only one', but 'every individual one'. I was asking about the difference between concluding "at least one is different from at least one other" and "profiles 1, 13 and 28 each differ from the average of all others beside themselves". (Other possibilities exist, such as wanting to identify clusters, so that you might say 'profiles 1,13 and 28 fall into one group and all other profiles form the main group'.) | |
| Apr 12, 2014 at 11:18 | comment | added | Py-ser | Sorry, I can't understand your last question. You mean: "identify only ONE profile that is different..." ? Could you explain it better? | |
| Apr 12, 2014 at 10:58 | comment | added | Glen_b | Yes, that's what I meant. Is your interest in identifying each one that's significantly different, or just that some are in fact different from others? | |
| Apr 12, 2014 at 10:31 | comment | added | Py-ser | You mean, every time I want to test a single profile, I should compare it with the average one obtained without including the single I want to test, right? | |
| Apr 12, 2014 at 10:26 | comment | added | Glen_b | Dependence may be an issue. At the least, you'd want to separate out the single profile from the average of the rest (which could be done by a linear combination of the overall average and the single profile), which would remove one form of dependence (though there might still be some time-dependence issues). | |
| Apr 12, 2014 at 7:20 | history | asked | Py-ser | CC BY-SA 3.0 |
