Skip to content

creating an equivalent of r-help on r.stackexchange.com ?

4 messages · Barry Rowlingson, Marc Schwartz, Gabor Grothendieck

#
As one of the original ranters of "hey lets move to StackOverflow" a
few years back (see my UseR! lightning talk from Warwick) I should
probably stick my oar in.

I don't think the SO model is a good model for all the discussions
that go on on R-help.

I think SO is a good model for questions that have fairly precise
answers that are demonstrably 'correct'.

I think a mailing list is a bad model for questions that have answers.
Reasons? Well, I see an email thread, start reading it, eight messages
in, somewhere in a mix of top-posted and bottom-posted content, I
discover the original poster has said "Yes thanks Rolf that works!".
Maybe I've learnt something in that process, but maybe I had the
answer too and I've just wasted my time reading that thread. With
StackOverflow questioners "accept" an answer and you needn't waste
time reading it. I've given up reading R-help messages with
interesting question titles if there's more than two contributors and
six messages, since its either wandered off-topic or been answered. I
suspect that heuristic is less efficient than SO's "answer accepted"
flag.

SO questions are tagged. I can look at only the ggplot-tagged
questions, or the 'spatial'-tagged questions, or ignore anything with
'finance' in it. Mailing lists are a bit coarse-grained and rigid for
that, and subject lines are often uninformative of the content.

SO is smart. Users are dumb, right? How many R-help questions could
have been answered by googling or reading the documentation? SO
compares input questions with existing questions, and suggets to users
that maybe this question here has the answer. How cool is that? And
the more questions and answers it has, the smarter that system gets.
Duplicate questions can be manually flagged by moderators.

SO questions get edited by other users, including fixing typos and
tagging properly. And bad questions are moderated out of existence, so
you don't even see them. How would you like to never see an R FAQ 7.31
question ever again?

For general discussion of R-related topics I think R-help is a better
place than SO but please don't make the mistake of thinking SO is just
another "web-forum" which those pesky kids on my lawn are promoting
instead of my cuddly old mailing list. Its a brilliant
question-and-answer *service*, which could not work as well as it does
over email.

 I also don't think a specialised R StackExchange site would be a good
idea either, since the site software is not suited to discussions and
the site would just fill with rambling guff.

 In summary: got an R programming question that you think has a
definite answer? Post to SO. Want to ask something for discussion,
like what options there are for doing XYZ in R, or why lm() is faster
than glm(), or why are these two numbers not equal - post to R-help.
Questions like that do get posted to SO, but we mod them down for
being off-topic and they disappear pretty quickly.

 Personally I still don't like mailing lists for discussions, but
StackExchange sites are not the place for discussion and I'm not sure
a better place exists that would keep everyone happy anyway!
On Mon, Feb 3, 2014 at 9:49 PM, Ted Harding <Ted.Harding at wlandres.net> wrote:
#
Hi All,

As I have noted in a prior reply in this thread, which began last November, I don't post in SO, but I do keep track of the traffic there via RSS feeds. However, the RSS feeds are primarily for new posts and do not seem to update with follow ups to the initial post.

I do wish that they would provide an e-mail interface, which would help to address some of the issues raised here today. They do provide notifications on comments to posts, as do many other online fora. However, there is no routine mailing of new posts with a given tag (eg. 'R'), at least as far as I can see, as I had searched there previously for that functionality. That would be a nice "push" based approach, as opposed to having to go to the web site.

I appreciate Don's comments regarding too many web site logins and too many passwords. Slight digression. The reality of constant security breaches of web sites has led me to use 1Password, such that I have a unique, randomly generated, strong password for almost every site that I login to (where I can control the password and login). I don't have to remember user IDs and passwords. With the multiple browser plug-ins for the application on the desktop and mobile app support with cross platform syncing, this has become, operationally, a non-issue for me.

I think that Barry makes a good distinction here. Notwithstanding the "gamification" of posting on SO, the formalisms on SO are pretty well ingrained.

I do also think that the "marketplace" (aka R users) in many respects, is speaking with its fingers, in that traffic on R-Help continues to decline.

I am attaching an updated PDF of the list traffic from 1997-2013, which at the time that I posted it last year, was not yet complete for 2013, albeit, my projection for the year was fairly close.

You can see that since the peak in 2010 of 41,048 posts for the year, traffic in 2013 declined to 20,538, or roughly a 50% decline. Much of that decline was from 2012 to 2013, which I postulate, is a direct outcome of the snowballing use of SO primarily.

Not in the plot for this year, January of 2014 had 1,129 posts, as compared to January of 2013 with 2,182 posts, or roughly a 50% decline. So the trend continues this year. If January's relative decline holds for the remainder of the year, or worse, perhaps accelerates, we could end the year at a level of activity (~10k posts) on R-Help not seen since circa 2002.

I honestly don't know the answer to the question and don't know that SO is the singular solution, as Barry has noted. However, as a long time member of the community, do feel that discussion of the future of these lists is warranted.

Perhaps Duncan's prophecy of R-Help just passively fading away will indeed happen. If the current rate of decline in posts here continues, it will become a self-fulfilling prophecy, or at minimum, R-Help will be supporting a declining minority of R users. Is it then worth the time, energy and costs to maintain and host, or are those resources better directed elsewhere to yield greater value to the community?

Should this simply continue to be a passive process as the marketplace moves elsewhere, or should there be a proactive discussion and plan put in place to modify infrastructure and behavior to retain traffic here? I suspect that this year may very well be important temporally to the implications for whatever decisions are made.

Regards,

Marc Schwartz







-------------- next part --------------
A non-text attachment was scrubbed...
Name: R-Help-Annual.pdf
Type: application/pdf
Size: 4854 bytes
Desc: not available
URL: <https://stat.ethz.ch/pipermail/r-help/attachments/20140203/467964e0/attachment.pdf>
-------------- next part --------------
On Feb 3, 2014, at 6:34 PM, Barry Rowlingson <b.rowlingson at lancaster.ac.uk> wrote:

            
#
On Mon, Feb 3, 2014 at 8:41 PM, Marc Schwartz <marc_schwartz at me.com> wrote:
You can set up email subscriptions for specific tags.  See the
preferences section of your account.  I get regular emails of the
r_filter.
Here are the first few lines of an email I juist received (I have
pasted it into this text plain email but they are received as HTML and
there are links to the specific questions).



159+ new questions in r filter on stackexchange.com

________________________________

R: read .dta file and use value labels only for selected variables to
create a factor

What is the easiest way to read a .dta file in R and convert only
specific variables as factors, using Stata value labels? I didn't find
a way to specify the convert.factors option in the foreign ...
Tagged: r stataby tfr on stackoverflow.com
________________________________

sort.col command in Splus to R

I have a code in Splus, but have to convert it into R, which is not a
big thing. However I am very new to both softwares. This is the code I
am struggling with: bestmodind <- ...
Tagged: r matrix s-plusby akeenlogician on stackoverflow.com
#
On Feb 3, 2014, at 8:54 PM, Gabor Grothendieck <ggrothendieck at gmail.com> wrote:

            
<snip>

Thanks for the pointer Gabor. I did not have an account on SE/SO and had only searched the various help resources there attempting to find out what kind of e-mail push functionality was available. A number of posts had suggested a "non real time" e-mail ability, which indeed seems to be the case.

I went ahead and created an account to get a sense of what was available. As you note, you can sign up for e-mail subscriptions based upon various tag criteria. However, it would seem that you need to specify time intervals for the frequency of the e-mails. These can be daily, every 3 hours or every 15 minutes. So there seems to be a polling/digest based process going on.

I created an e-mail subscription last evening and selected every 15 minutes. What appears to be happening is that the frequency of the e-mails actually varies. Overnight and this morning, I have e-mails coming in every 20 to 30 minutes or more apart. It is not entirely clear what the trigger is, given the inconsistency in frequency. Perhaps the infrastructure is not robust enough to support a more consistent polling/digest e-mail capability yet.

The e-mails contain snippets of new questions only and not responses (paralleling the RSS feed content). I need to actually go to the web site to see the full content of the question and to see if the question has been answered. In most cases, by the time that I get to the site, even right away after getting the e-mail, there are numerous replies already present. There is, of course, no way to respond via e-mail.

I would say that if one is looking for an efficient e-mail based interface to SE/SO, it does not exist at present. It is really designed as a web site only interaction, where you are likely going to need to have a browser continuously open to the respective site or sites in order to be able to interact effectively, if it is your intent to monitor and to respond in a timely fashion to queries. 

Alternatively, perhaps a real-time or near real-time updating RSS feed reader might make more sense for the timeliness of knowing about new questions. It is not clear to me how those who respond quickly (eg. within minutes) are interacting otherwise.

There appear to be some browser extensions to support notifications (eg. for Chrome), but again, you need to have your browser open. There also appear to be some desktop apps in alpha/beta stages that might be helpful. However, they seem to track new comments to questions that are specifically being followed (eg. questions that you have posted), rather than all new questions, thus paralleling the SE/SO Inbox content.

That being said, obviously, a lot of people are moving in that direction given the traffic decline here and the commensurate increase there.

Regards,

Marc