Rob Hyndman is trying to start a question and answer site for people working in Statistics, running on the same software system that supports our beloved MathOverflow. Due to a change in the way StackExchange is licensing their platform, he can’t simply get up and start using it. Rather, he must first obtain a critical number of users who will commit to regularly checking in and using the site. If you would be interested and capable of making this commitment, head on over to Statistical Analysis — Area 51. I find the new policy strange, but that’s apparently how things are working now.

I think that a Statistics version of MathOverflow would be a great idea. All the time, people come to me or to MathOverflow asking questions that are really meant for a statistician. There are a lot of sophisticated questions in Statistics, and it would be good to have a place for them to be answered.

I also think it would be good for us mathematicians to check in more on what the statisticians are doing. Great math has often been driven by the needs of the other sciences; and one of the big needs of today is for better methods for analyzing large data collections. I can say that my advisor, Bernd Sturmfels, has found a lot of pretty algebraic geometry problems by trying to help out our local statisticians.

Disclaimer: I don’t know Professor Hyndman, and have no involvement with this project other than wanting it to succeed.

### Like this:

Like Loading...

*Related*

There’s also a proposal for a StackExchange site on LaTeX. It needs more people to come and help select a list of canonical “on-topic” and “off-topic” questions to define the scope of the site. After that it will proceed to the stage the Statistical Analysis proposal is it, asking for commitments from potential participants.

(For the record, I also think that the Area 51 process is pretty strange, and not that likely to work, but nevertheless I’d really like it to.)

I must be missing something obvious, but why not create a

Statisticslabel on MathOverflow… and discuss Statistics on MO? After all, isn’t Statistics a field of Mathematics? I find this “neo-tribalism” counter-productive. There are deep connections between Statistics and the rest of Mathematics, so why promote isolationism?Hello David,

Thank you for supporting this project, I am a strong believer in it as well.

I also wrote about it on my own blog ( R-statistics.

To answer Rod’s question –

It would be like asking why not to have a physics label on mathoverflow.

Statisticians deal a lot with data cleaning, visualization, model fitting, programming, work flow and the likes.

And community wise, there is a very distinct “data analysts/ machine learning / statisticians” community – different then that of the mathematicians.

Best regards,

Tal

Tal,

I see your point. I thought that the goal was to discuss

Mathematical Statistics, not to deal with data analysis, programming, and models. Having said that, I now think it’s a great idea!If you look at the MO questions tagged st.statistics I think it’s pretty clear that there are a lot of interesting statistics questions that would fit better with their own site. For example, this question is perfectly at home on mathoverflow, while this question is an interesting question, but not a great fit for MO.

@Rod: I would say that statistics is a field of mathematics to a similar extent that electrical engineering is a field of physics.

Mark,

Statistics has one foot in solid Mathematics, and the other foot on some “voodoo art” that involves dealing with data. I have been following this blog since its inception, and it always focused on the purist, mathematical side of things. I assumed that by “Statistics” the author meant “Mathematical Statistics”… not the voodoo stuff. My mistake. I have been corrected. Time to move on.

Even leaving aside data collection and analysis, the fact is that people who call themselves mathematicians are seldom well equipped to answer sophisticated questions about mathematical statistics. Just like many physicists aren’t well equipped to answer many physics questions that arise in electrical engineering.

Mark,

In my humble opinion, it’s a lot worse than that. I know analysts who aren’t well-equipped to answer not-too-sophisticated questions about Geometry, Algebra, or Combinatorics, etc. I am an EE and I certainly am utterly ignorant about most fields within EE. Let’s face the harsh reality: we, mere mortals, can’t aspire to be much more than

fachidioten. Humans have created too much knowledge in the past 200 years for one to keep up.Comment number 10 is spam

Killed it, thanks.

Now your comment looks like an interesting piece of self-referential spam.

lol – feel free to erase my old (and the current) comments :)

Are you going to call it StatOverflow?? Seriously, I have lots of questions about the Gaussian Unitary Ensemble when the site is up.

John: depending on exactly what your questions are, MathOverflow may be a better place for questions about the GUE. (Not necessarily the best place, but there are a few people on MO who know a lot about the GUE who aren’t necessarily likely to be around a stats site.)

John: also, personally I’d love to see the average quality of probability questions on MO go up, so please, ask away!

This site already exists:

http://metaoptimize.com/qa/

I haven’t announced it widely yet, but it’s for statistics, machine learning, NLP, and other adjacent fields.

I am focusing on getting buy-in from the academic community (something the Area 51 proposal doesn’t have) before announcing it more broadly.

As we’ve learned from StackOverflow, it makes sense to use a broad topic, rather than something too niche (i.e. just statistics).

And, unlike StackExchange, the OSQA platform is completely open source!

Thanks Joseph! This looks great! When can I start pointing people to it?

David, Feel free to announce it on your blog and point people to it!

I’ll send you an announcement email you can use.

Now the site has reached 100% commitment. Beta will start soon.