As others see us: "vitriolic hatred"

[Nick Cox]

I don't ever post on the most used social media, or try to follow any systematically, but I occasionally stumble against mentions of Statalist elsewhere. Here is Dr Claire Wylie on Twitter:

Why are the people in statistical help forums so rude, releasing a vitriolic hatred towards us less statistically blessed souls who are only on there because we don’t know how to ask our question otherwise we’d have googled it! #stata #rstats #statalist #stackexchange #justbekind

https://twitter.com/DrCEWylie/status/997242488875241472

This kind of post raises reflections on various levels, chiefly

1. Really "vitriolic hatred"? Examples please?

2. A technical forum inevitably involves back and forth when questions seem unclear and confused. We do try to provide detailed, constructive guidance, much of which is ignored every day! Should we just ignore all imperfect questions?

3. If anyone sincerely detects anything of the sort, we should be talking about it. Or you can always bring posts you disapprove of to the attention of the administrators at StataCorp (mine too, naturally).

4. Anyone lurking who wants to contribute and be nicer, faster, better at answering questions here is welcome on any or all of those scores.

Claire doesn't appear to be a member -- if she has posted here, she used another name -- but like anybody else she is entitled to read and disapprove of anything she dislikes. FWIW, I am puzzled that angry, blanket denunciations of entire communities go together with exhortations to "just be kind".

[Andrew Lover]

I wonder if Statalist was simply lumped in with stackexchange and rstats (both of which can at times have the issues highlighted)?

[Nick Cox]

I post on Stack Overflow and Cross Validated too. I don't want to discuss them particularly here, but anyone can broaden the question. I don't find a systematic difference in what I read (or post...). An important difference for SO and CV is active moderation, so that moderators can delete comments, suspend people, etc. (Since Statalist went to a forum in 2014 to my knowledge several spammers and just one other person have been removed by StataCorp.)

#rstats in context could mean the Twitter flag. R-help is fast fading from its peak, so far as I can see. Although it often got mentioned as unfriendly, its terms of reference are different. Any way, I've never posted under either heading.

My concern is here with Stata and Statalist, explicitly flagged.

[Richard Williams]

Probably the most seemingly-negative comments are about people not following the FAQ guidelines on posting questions effectively. Or, berating posters when they fail to follow advice previously given and keep posting anyway. I view such comments as constructively telling people what they need to do to get an answer, but maybe some take offense at that. In general we seem pretty civil to me, but I suppose there are exceptions. One thing about online posting is that things might be taken negatively when no offense was intended. And, even if posters are chided, they usually get an answer if someone knows the answer.

I also think our preference for real names discourages obnoxiousness. I know it is one of the things I like about Statalist. If I am helping someone I like them to know who it is, and if I am helped I like knowing who to thank. (The only bad thing is that when people see me on Statalist, they sometimes email me directly, often about statistical methods I know nothing about!)

[Weiwen Ng]

I don't see anything I would count as vitriolic hatred. There was one time one guy posted something, then posted back something along the lines of "thanks for nothing" a few hours later when he resolved his problem but didn't get any responses in the meantime... that's the worst behavior I can recall.

There are definitely a few posters here who could come across as being persnickety, but that's generally been for the (IMO valid) reason that it is often difficult or impossible to help someone if we have no clue about their data structure.

There is one potential problem with the strong recommendation to post a data sample. I work in health care, and we often have clinical or interview data from patients. I don't actually know if HIPAA (a very long US law that governs privacy of medical data) prohibits me from posting a de-identified sample of, say, survey responses (e.g. list a subset of survey responses, no identifiers at all, no other variables attached, for the first 100 people in my dataset). However, I've generally assumed that I can't. I've been able to get much of the help I've needed without that, but I have fairly good programming skills, and I believe I've seen cases where people have implied that they think they can't post data (then again, there was this one guy who posted his data and patient names...). If I had weaker programming knowledge, it might be tricky to know that you can, in fact, subset your data and variables in -dataex-, and it might be even trickier to adapt some stock Stata dataset to illustrate your question, or to generate random data to illustrate it.

I don't know what to do about the above, however. I would like to see some sort of exception carved out in the HIPAA regulations, or by modifying the statute, that governs posting a subset of data for technical assistance in public fora. But the number of people who might benefit from this may not be very large. I don't see lawyers being comfortable with this, either, for fear that someone might make a mistake and accidentally post information that could be construed as potentially identifiable.

[Nick Cox]

We don't want anyone to embarrass themselves or others by posting confidential or sensitive data. The FAQ Advice already includes the comment that fake data are OK. I don't know what else could be helpful but still simple and not potentially problematic.

Adding a randomiser of some kind of dataex is tricky and might easily destroy the problem the data are designed to illustrate.

[Clyde Schechter]

I'd like to amplify a bit on Weiwen's remarks pertaining to HIPAA.

Depending on the source of the data, it can be a HIPAA violation to post example data that contains protected health information. You can, of course, remove identifiers (names, medical record numbers, etc.) and either omit them entirely from the example if they are not relevant, or replace them with arbitrary numbers or character sequences. You can also add some jitter to the other variables so that the data are realistic but not real. Similarly, for most kinds of questions, the names of the variables can be disguised. Of course, for the earliest beginners, just doing these things might exceed their skills.

Perhaps we should stress that what we need is an example that properly reflects the organization of the data, storage types, and labeling, and it should contain observations that illustrate the problem that needs to be solved.

Also, if you are working with HIPAA-covered research data and need help with the analysis, it is also a HIPAA violation to allow the data to be seen by a colleague who is not registered as a co-investigator on the project with your institutional review board. So this issue is, in theory at least, no different on Statalist than when seeking face-to-face help at one's own institution.

Concerning the original topic of this thread, while people (including me) sometimes get a bit testy here, in my opinion the tone has been quite professional nearly all of the time. I do wonder what Dr. Wylie is thinking of when she refers to "vitriolic hatred," as I cannot recall anything that comes close to meeting that description. Then again, such things are, I suppose, in the mind of the beholder, and in these times we seem to be seeing more and more people with extremely thin skin.

Added: Crossed with #6. The idea of adding a randomizer to -dataex- has crossed my mind from time to time, and I even developed a little wrapper for -dataex- to do it, but some experimentation with it quickly made it clear to me that, as Nick opines, it would too frequently destroy relationships in the data that are germane to the question being posed.

[John Mullahy]

FYI, the original tweet now has a response thread prompted by a tweet by Weiwen Ng. https://twitter.com/weiwentg/status/999387374164631564

[Claire Wylie]

Dear all,

Let me apologise for my somewhat hyperbolic blanket denunciation tweet, and in doing so clarify it.

Firstly, Andrew is correct - the hashtags were generic. #Statalist is populated if you #Stata - and my intention with the hashtags was to lump together my thoughts. I had previously posted a query on Stack Exchange / Cross validated regarding the logical selection of variables based on p-values or size of effect within Stata for a very large forward stepwise logistic regression. The response was that I was stupid and deserved to fail my project. Both these points may be correct - but I didn't ask my query to feel worse about myself - hence it prompted my tweet.

Secondly, Nick is correct - I was not a member of Statalist and it is therefore unfair that my tweet has suggested that its members are likely to respond in the same way. He is also correct that I was angry and should have waited before phrasing my tweet more appropriately.

Thirdly, it is encouraging that the tweet has prompted this forum to reflect on best practice - whether or not that was even necessary. I pick up on the comment of 'extremely thin skin' and think this may merit more empathy. Personally, I try very hard to answer my own problems - usually spending days on end Googling phrases or words that may reflect what I'm trying to do - but often I'm not really sure what I'm asking - just that I know I'm not achieving the correct result, or that what I'm doing is wrong. I imagine that many other people who then resort to using forums - such as Cross Validated or this one or the like, may also be as exasperated, frustrated and despondent as I am by the stage they try to formalise their question. Sometimes this level of stress may mean that responses such as 'I can Google this for you', curt/blunt responses, or downright rudeness (in my logistic regression example) can perhaps be the straw that breaks the camel's back. In learning from this tweet's experience, perhaps in these situations it is best if we have nothing nice to say, to say nothing at all.

Finally, I did not intend to bad mouth this forum, hence I will delete my tweet as I wouldn't want others to reach for #statalist and be put off using it. From what I can see so far, the advice looks professional and useful, and I certainly hope to be able to come here as my first port of call for my Stata queries from now on. Perhaps if I do so before spending many days Googling my problems I'll reduce my stress levels somewhat.

Thank you for reflecting on my tweet, and apologies for the unnecessary upset it caused.

Claire

[Bruce Weaver]

Welcome aboard, Claire. Now let's all have a cuppa tea...







...and get back to analyzing data.

[Clyde Schechter]

Claire Wylie Welcome to Statalist! While the culture here is to be frank and direct when people make mistakes, it is also the culture to do it in a professional way, and certainly to avoid anything demeaning (in the non-statistical sense of the word.) And it is also the culture here to acknowledge and accept criticism offered publicly, and to publicly acknowledge and correct our errors. We strive to be a learning community. Based on your thoughtful post in #9, I predict you will fit in rather well here, and will probably find it a more congenial environment for seeking help with Stata.

To my knowledge, nobody here has ever called anyone stupid or said they deserve to fail. Even the most experienced members of the Forum were once beginners, and we all grappled with learning Stata and statistics, and, I think, we all remember those days. Your comment in #9 about the need for empathy is a good one, and I, for one, will take it to heart. Your examples of context are good reminders. My remark about thin skin was meant generically, and to the extent it was aimed at you it is because I had mistakenly assumed that your complaint about vitriolic hate arose from some interaction with Statalist. Of course, had I examined your tweet more closely and seen the array of hashtags, I might have avoided that assumption--I will try to exercise more care with that.

hence I will delete my tweet as I wouldn't want others to reach for #statalist and be put off using it

I have never used Twitter, and my only familiarity with it is when some web site that I frequent links to it. I gather it is common and accepted practice there to delete tweets when one has second thoughts. Here the culture is different, and we ask people not to delete their posts, even if they have solved their own problem. This is because there are people who never post, but who read what others post so that they can learn from the questions and answers. (I did this myself for many years.) So if people find their own answers, rather than deleting the post, we prefer that they add a new post showing the solution.

I look forward to working with you here.

[Nick Cox]

Time zones alone imply that Bruce and Clyde have already said most of what should be said, and said it very well, in thanking Claire for her very helpful explanation and discussion and apology. I add my own personal thanks.

Specifically I don't recall and can't find the post on CV that led to a comment about being stupid and deserving to fail a project. I can confirm as a member of CV that I would flag that kind of comment for moderators and I am confident that, at a minimum, it would been have removed. Similarly if such wording were used here on Statalist I very much doubt that it would have passed unchallenged. It's not so long ago that an all too common swear word was edited out!

[Carlo Lazzaro]

Claire:
welcome on this forum and thanks for your clarifications.
It's true that -stepwise- is not very sponsored on this forum, too.
However, as far as I can remember, no poster has been ever blamed for asking how to perform it, but was instead referred to a methodological list that explains why that approach should not be recommended (https://www.stata.com/support/faqs/s...sion-problems/).
I do hope you'll enjoy the list.

[Weiwen Ng]

...
Thirdly, it is encouraging that the tweet has prompted this forum to reflect on best practice - whether or not that was even necessary. I pick up on the comment of 'extremely thin skin' and think this may merit more empathy. Personally, I try very hard to answer my own problems - usually spending days on end Googling phrases or words that may reflect what I'm trying to do - but often I'm not really sure what I'm asking - just that I know I'm not achieving the correct result, or that what I'm doing is wrong. I imagine that many other people who then resort to using forums - such as Cross Validated or this one or the like, may also be as exasperated, frustrated and despondent as I am by the stage they try to formalise their question. Sometimes this level of stress may mean that responses such as 'I can Google this for you', curt/blunt responses, or downright rudeness (in my logistic regression example) can perhaps be the straw that breaks the camel's back. In learning from this tweet's experience, perhaps in these situations it is best if we have nothing nice to say, to say nothing at all...

This is a good point. There is a bit of a theory of mind issue: when I started working with statistical software, I was in SAS, and I was trying to figure out how to change my data from long to wide. After some fumbling in Google, I found out that you would use PROC TRANSPOSE in SAS (all caps is their writing convention, not mine), but that you'd use -reshape- in Stata. By that time, my programming skills were intermediate-ish. I can easily see how beginning programmers would not quite know what to Google.

I believe I've generally kept in mind that knowledge gap when responding to people. This is a good reminder to keep doing so.

It's been my observation that programming or data manipulation skills are distinct from statistical skill. They're likely correlated, but they are separate concepts. Here, the usual responders tend to have both, which likely stems from selection bias. In my work, my principal investigator (PI) directly hires one guy for the majority of the data manipulation, and in both my previous jobs, the organizations did the same. Ultimately, though, many statisticians have to do some work themselves or have RAs whose primary knowledge base isn't data manipulation to do it. It would be nice if we could teach data manipulation more systematically in PhD programs, but I don't perceive that there's the capacity to implement it - both in terms of student time and availability of teachers. So, again, I'm not quite sure what to do with this, and it could just be one of those things we have to live with.

I shall wrap up by saying that I have found the Stata forum to be a great place to learn. At the very worst, people sometimes come across as persnickety, but that often stems from the fact that we don't know what your data look like, or that we are trying to keep the place organized enough that people can search for threads that might help them. I will strive to keep this behavior on my part within the bounds of something that a reasonable person can not take personally.

[Doug Hemken]

Some days I feel impatient with the-same-old-question or the-poorly-expressed-question. Those days I tell myself, "let someone else answer it."