We all acknowledge that statistics are a core tool for data scientists. But considering that Cross Validated exists to answer questions specifically concerning statistics, what type of statistics questions (if any) remain within the scope of this site?
Here are some examples of statistics questions that have been asked here so far:
The thread "What characterises the difference between data science and statistics?" is relevant here.
My take: as in the thread linked above, statistics are a part of data science... together with hacking skills and domain expertise. I believe that specific questions about statistics that do not connect with non-statistical data mining or software or domain specifics would be better at CrossValidated, for two reasons:
Statistics questions are more easily answered by statisticians... who hang out at CrossValidated, not here.
Answered statistics questions may be helpful for other people in the future. But those people will look for answers to their stats questions on CrossValidated, not here.
Similarly, I would argue that questions about installing a specific piece of data science software would be better at SuperUser, questions about how to do something in R in the R tag at StackOverflow, and domain specific questions about a data science project in chemistry at Chemistry.
Yes, there will be a lot of gray areas. But it seems to me like the community has done a good job in migrating the first and third of the questions linked to in the Q to CV, while the second and fourth question stayed here.
Data science is a cross disciplinary field. Questions about application of statistical methods for various real life datasets processing should be on topics.
The questions that are given as examples here have very little connection with applied aspect of statistical analysis for real life datasets. By real life datasets I mean data generated and collected from daily life, like web traffic data, economic activity, open data programs datasets etc. as opposed to lab generated data or experiments statistical evaluation.
The example questions presented here are too abstract, doesn't fall into intersection between statistics and programming, and should be classified as off topic.
Some thoughts: I believe that overlap with CV exists and should exist. We are building a community, and I believe we should encourage people put their non-theoretic questions on stats on this site.
I am the author of the first question above and my question was a very real situation I have to deal with in my company. Of course I could ask it in CV, however, I believe that it would be of interest for the members of this site. Probably the same opinion had the at least 5 people that upvoted it, the 2 people that answered it and the 14 people that upvoted their answers (duplicates included).
Similar scenario for question on p-values. The big-data concept and all the implications on statistics are something that statisticians are only very recently start to have to deal with (and history will show how ready they are to change mindsets - so many fields have originated in stats and quickly became independent because classic statisticians were not able to follow or adapt).
Overall I believe we should be open to most statistical questions rather than be ready to declare them off-topic. I prefer to go big and host the union of the different fields/topics that comprise "Data Science" rather than their small intersection.
