Area 51 » Science

218

followers

Announcements

I'd like to propose a change to 'Proposed Q&A site for statistics, data analysis and data mining'. It sounds a bit trivial, but the name as it stand implies that only people who do only statistics for a living should use the site. There are people in many other professions who might have a lot to contribute to the site, but be put off if they feel like it is just for people where statistics is their profession rather than part of their profession. – Ian Turner May 27 '10 at 20:57

Statistics is not a subset of mathematics. It uses some mathematics, but so does physics, engineering, etc. Try asking any of these questions on MO and they will (rightly) be closed as off topic. – Rob Hyndman May 27 '10 at 23:40

I've modified the subheading to take account of the comments being made. Let me know if you think this is better (perhaps vote up this comment). I'm still not sure what to do with the major heading. I understand the objections, but haven't come up with anything better yet. Feel free to make some suggestions in the comments. – Rob Hyndman Jun 1 '10 at 3:55

show 14 more announcements

113 Example Questions (12 closed)

active oldest votes

up vote 2 down vote

added by therefromhere May 28 '10 at 16:18

up vote 2 down vote

added by Joe Mako May 29 '10 at 20:20

An example would be if you have the average occurrence of something for each state, why does an average of an average not make sense to get the country average. Or is there any value in looking at the median of a set of averages? – Joe Mako May 29 '10 at 20:22

up vote 2 down vote

added by Mehper C. Palavuzlar Jun 3 '10 at 10:32

you can search this term in r-project.org – maahtu Jun 6 '10 at 13:25

up vote 2 down vote

added by Mehper C. Palavuzlar Jun 3 '10 at 10:41

up vote 2 down vote

added by Menno Jun 5 '10 at 7:41

up vote 1 down vote

added by Rob Hyndman May 25 '10 at 2:09

up vote 1 down vote

added by Rob Hyndman May 25 '10 at 2:11

Questions like this often end up on StackOverflow, but I think this site should cover tool-based questions such as this. – Rob Hyndman May 25 '10 at 12:41

It feels rather weird allowing question on R packages, but refusing R code/implementations. As with Math Overflow, it might not be so bad to allow code snippets or these grey area topics. Especially when SO might not want this question – Ivo Flipse May 26 '10 at 21:31

I think this site has got to allow some implementation/tools discussion/answers in order to have practical relevance. Unless you're going to calculate it by hand, it doesn't do much good to know how a logistic regression works without having tools to implement. However, the trick will be in only allowing tools discussions that are on-topic. For example, how to use SAS to connect to a Sybase data warehouse is likely off-topic even if you want to run a logistic regression on the data you get. – Jay Stevens May 27 '10 at 10:29

I think there is a difference between asking for code vs asking for existence of a package. – Anon Jun 2 '10 at 3:01

(Is there am R package for SVMs? It would help me a lot!) – Larry OBrien Jun 4 '10 at 17:25

up vote 1 down vote

added by Joel Spolsky♦ May 25 '10 at 3:16

closed as off topic by Rob Hyndman, Mike Grady, Jeromy Anglim, Yacoby, dswan Jun 6 '10 at 10:47

This question does not relate to the topic of the proposal.
It was closed as part of an automated migration of off-topic to close votes on September 29, 2011.

too easy, I assume, and sounds like homework. – Joel Spolsky♦ May 25 '10 at 3:17

I'd rather not see any questions that ask others to do computations for the asker. How to compute, yes; result of computation, no. – Matt Parker May 27 '10 at 18:26

I agree with Matt! Concepts should be clarified, method on how solving it too. But no answers. No homework :P – George May 28 '10 at 4:47

No answers for simple calculations such as this - I agree with that. However, what about checking with stats professionals about complex answers or confirming the implementation of a stats routine in a new language / program? – Patrick Welch Jun 4 '10 at 20:38

Patrick - I'm all for that, and I think questions like that will set themselves apart from this one because of the complexity and because the asker will clearly have put some substantial effort into their own answer. – Matt Parker Jun 8 '10 at 1:34

up vote 1 down vote

added by Rob Hyndman May 25 '10 at 5:52

closed as off topic by Matt Parker, Mike Grady, dswan, Karsten W., Joe Mako Jun 7 '10 at 4:56

This question does not relate to the topic of the proposal.
It was closed as part of an automated migration of off-topic to close votes on September 29, 2011.

Best to leave questions about R code on StackOverflow. – Rob Hyndman May 25 '10 at 5:52

@Rob: I disagree. Just because a question is on-topic on another site doesn't mean that it's off-topic on every other site. If there were a SE site dedicated to statistics, I think it would be pretty artificial to exclude R questions. If a professional statistician would ask this question (in her capacity as a statistician), why isn't it on-topic? – Anton Geraschenko May 27 '10 at 6:42

@Anton. The trouble is, the site could quickly degenerate into general tech support for every tool being used. There are already several places to go for R tech support including SO, R-help mailing list, etc. Similarly for SAS, SPSS, etc. There is nowhere to go for good advice about statistical methodology, data analysis, etc. – Rob Hyndman May 27 '10 at 7:41

Another point to consider here is that this question is not about statistics at all, it is simply about R syntax. A question about how to do something statistical in R (e.g., how to fit a mixed-effects model) would be appropriate. To test the waters, I'll add such a question and see how it fares. – Rob Hyndman May 27 '10 at 12:23

@Rob: Thanks for clarifying. I agree with you now. – Anton Geraschenko May 27 '10 at 20:10

show 2 more comments

up vote 1 down vote

added by Anton Geraschenko May 27 '10 at 7:26

up vote 1 down vote

added by WikiSpeedia hang-around May 27 '10 at 12:16

(From en.wikipedia.org/wiki/Mahalanobis_distance: "Mahalanobis distance is widely used in cluster analysis and classification techniques. It is closely related to Hotelling's T-square distribution used for multivariate statistical testing and Fisher's Linear Discriminant Analysis that is used for supervised classification.") – WikiSpeedia hang-around May 27 '10 at 12:19

up vote 1 down vote

added by Ian Turner May 27 '10 at 12:17

up vote 1 down vote

added by Andrew Song May 27 '10 at 21:18

The R tag on stackoverflow is already pretty active: is the goal to move that community over to this site instead? – Shane Jun 1 '10 at 15:50

It does not have to be one or the other. Overlap is going to happen with the plethora of sites that will spring from Area51. – Stu Thompson Jun 1 '10 at 16:34

I agree with Stu that there will be some overlap. However, as a general rule I think programming and syntax questions should go to StackOverflow and statistical questions should go here. The boundary is obviously fuzzy. – Rob Hyndman Jun 1 '10 at 22:09

up vote 1 down vote

added by Jay Stevens May 29 '10 at 12:24

up vote 1 down vote

added by chollida May 31 '10 at 17:25

en.wikipedia.org/wiki/Kurtosis ... so I'm inclined to say it's not a good example. – Matt Parker Jun 8 '10 at 17:04

@Matt Parker, all you did was post a link to a wikipedia article. It's entirely unclear what point you were trying to make with the link. Can you expand on your reasoning? – chollida Jun 9 '10 at 19:28

My apologies. The point I so feebly tried to make is that, while on-topic, the question isn't one that will make make experts think, "Finally, this is a statistics site I can get something from." It's just too basic to serve as an outstanding example of what the site should be about. If the question were more like, "I'm fitting a linear regression to a dependent variable with high kurtosis. How will this affect the model?", I think it would suit the aims of the StackExchange sites better - i.e., to create knowledge-sharing communities of experts. – Matt Parker Jun 9 '10 at 23:06

@Matt Parker, great response Matt, I agree with you. – chollida Jun 10 '10 at 17:02

up vote 1 down vote

added by Anon Jun 2 '10 at 3:22

up vote 1 down vote

added by Foster Brereton Jun 2 '10 at 20:48

up vote 1 down vote

added by Mehper C. Palavuzlar Jun 3 '10 at 10:38

up vote 1 down vote

added by Karsten W. Jun 12 '10 at 11:57

up vote 0 down vote

added by Rob Hyndman May 25 '10 at 2:18

closed as off topic by Ian Turner, Mike Mazur, Shane, François G. Dorais, Michel de Ruiter Jun 15 '10 at 20:00

This question does not relate to the topic of the proposal.
It was closed as part of an automated migration of off-topic to close votes on September 29, 2011.

NOT BASEBALL STATS! I GET IT! haha – Joel Spolsky♦ May 25 '10 at 3:16

up vote 0 down vote

added by Rob Hyndman May 25 '10 at 2:21

	Should I use a three-dimensional bar chart, stacked bar chart, side-by-side barchart, or something else? – Rob Hyndman May 25 '10 at 2:21
	I voted yes, because I believe that data visualization is just as important as the actual data analysis. But, I am assuming that you will be presenting some information in the question about what the data looks like and what you are trying to analyze. – Waleed Al-Balooshi May 27 '10 at 12:27
	Data visualization is part of data analysis. In general, I am assuming that there will usually be more information than what is given in these short questions. – Rob Hyndman May 27 '10 at 12:32

up vote 0 down vote

added by Rob Hyndman May 25 '10 at 5:46

up vote 0 down vote

added by Harvey Motulsky May 25 '10 at 11:35

up vote 0 down vote

added by Ciaran De Buitlear1 May 25 '10 at 15:19

Best of lists should not be encouraged, but this is likely to be something A LOT of visitors might be interested in. – Ivo Flipse May 25 '10 at 20:59

It's on topic, but is subjective and argumentative. The body of the question might remedy this by explaining what problem the asker is trying to solve (i.e. best for what?). – Anton Geraschenko May 27 '10 at 6:48

This seems more suited for a community wiki type of questions. I think it needs to be rephrased. Maybe "What tool is most used in Econometric Analysis and Why?" – Waleed Al-Balooshi May 27 '10 at 12:21

voted up, but subjective no? – George May 28 '10 at 4:48

Subjective, unless Community Wiki. Compare stackoverflow.com/questions/1711/… – György Andrasek Jun 3 '10 at 2:49

up vote 0 down vote

added by Rog May 27 '10 at 4:57

	Such a common problem that people don't understand. Really improtant to have a good resource for this sort of thing. – Ian Turner May 27 '10 at 13:35
	I agree that it's important to have a good resource for it, but I just don't want to see a thousand "How do I graph Plot X in Software Y?" questions. They are essentially limitless and are probably better answered in the program's help files. – Matt Parker Jun 2 '10 at 23:55
	You will get an answer on SO for that – Karsten W. Jun 6 '10 at 15:17

up vote 0 down vote

added by Juan Antonio, edited by Rob Hyndman Jun 2 '10 at 4:18

closed as off topic by Yacoby, Karsten W., therefromhere, Joe Mako, Charles Stewart Jun 7 '10 at 12:35

This question does not relate to the topic of the proposal.
It was closed as part of an automated migration of off-topic to close votes on September 29, 2011.

Appropriate for SO but not for this site I think. – Anon Jun 2 '10 at 3:02

*scrape data...? – d03boy Jun 2 '10 at 4:06

d03boy - let's say there's a government website with hundreds and hundreds of tables of data that I want. If the tables and pages have a pretty regular structure, you can write a program to cycle through all of the pages, collect the data, and create a database on your local file. – Matt Parker Jun 16 '10 at 23:42

up vote 0 down vote

added by Stu Thompson May 27 '10 at 6:30

closed as off topic by Ian Turner, Yacoby, walkytalky, Harlan, Sinan Ünür Jun 16 '10 at 22:13

This question does not relate to the topic of the proposal.
It was closed as part of an automated migration of off-topic to close votes on September 29, 2011.

Do you mean Excel 2007 or 2010? – Waleed Al-Balooshi May 27 '10 at 12:35

2008 is the Mac version and it is actually different enough to the Windows version to cause this sort of problem. I would argue this question should be better sited on a Microsoft Excel Stack Exchange. – Ian Turner May 27 '10 at 13:36

Yup, I meant Mac. Clarified. – Stu Thompson May 27 '10 at 15:33

I agree with Splash. I really don't want this to be a MS Excel plots help site. There are plenty of resources on the web to help people with Office products, but precious few that are much use for statistics. – Matt Parker Jun 2 '10 at 23:59