Timeline for What measure can I use to select a number from a dataset which future values will most likely be closest to?
Current License: CC BY-SA 4.0
12 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jul 6, 2021 at 19:08 | answer | added | user115931 | timeline score: 0 | |
| Jul 1, 2021 at 18:04 | comment | added | Adrian Keister | Well, I'm happy to try to improve my answer if it's not quite what you were seeking. | |
| Jul 1, 2021 at 17:20 | comment | added | Ryan Jarvis | @AdrianKeister Based on your comment and answer, I think you got what I was actually looking for even if I didn't articulate my question perfectly. Upon reflection, I do believe I am actually looking for a mode, within a multimodal distribution, that meets certain criteria. Now I am trying to decide if that warrants a separate question. | |
| Jun 30, 2021 at 6:45 | comment | added | Vincent | If you consider the mode of your dataset an outlier, then your data is not clean enough. You will have to manually clean it first. | |
| Jun 30, 2021 at 1:08 | history | became hot network question | |||
| Jun 29, 2021 at 18:00 | history | tweeted | twitter.com/StackStats/status/1409934709929488385 | ||
| Jun 29, 2021 at 17:41 | answer | added | Adrian Keister | timeline score: 4 | |
| Jun 29, 2021 at 17:36 | comment | added | Adrian Keister | $15$ is not really an outlier. While some people define the term "outlier" differently, one common way to find outliers is the $75\%+1.5\text{IQR}$ rule, which in your case is $31.5,$ well above your maximum. I would say that what you really have is a bimodal distribution. One way to look at your problem is to find the modes ($4$ occurs almost as often as $15$), and then see which mode is "closest" (you have to define that term) to the median. Finding the modes in the first place might work well with a peak detection algorithm. | |
| Jun 29, 2021 at 17:31 | answer | added | Stephan Kolassa | timeline score: 20 | |
| Jun 29, 2021 at 17:27 | comment | added | David L Thiessen | "Closest" in what sense? In statistics we normally talk about closeness in the sense of the squared error, (y_observed - y_predicted)^2. Then we consider all the possibilities and the probabilities of those happening to find the mean squared error, and we try to find the estimate that minimizes that error. But from your description it sounds like you have a different idea, since you rejected "5" and "7.5" because those exact values didn't happen very often. | |
| Jun 29, 2021 at 17:08 | review | First posts | |||
| Jun 29, 2021 at 17:26 | |||||
| Jun 29, 2021 at 17:05 | history | asked | Ryan Jarvis | CC BY-SA 4.0 |