I am sorry to interrupt your meta discussion, but I am looking really forward to solve what's bothering me. Does anyone has experience with model-agnostics for variable importance? I am wondering if you can use them to discover overfitting.

@J.Doe Quoting @gung from earlier: We prefer substantive questions are asked on the main site [...]. There we have better facilities for answering (eg, formatting). Chat is for gossip & grumbles.

@mkt oh ok I see. I've posted also a question on that stats.stackexchange.com/questions/422451/…

Thanks @mkt

Hey guys, I'm wondering if the question I've got in my head would be a suitable one for this site. Basically, the context is that I'm working with a task system where parent tasks delegate work to child tasks to accomplish something. The parent tasks tend to run and complete in a handful of milliseconds but the child tasks can take substantially longer, in the realm of seconds. However, we humans tend to understand a parent being done when all its children are done.

With the database we're using, I know I can get aggregate statistics of count, mean duration, and standard deviation for task A (parent) and task B (child) individually. What I'm curious about is how to combine these numbers into a "decent" estimator of the mean and standard deviation of the time between a parent task starting and all its children being done.

Is this extended argument on meta about the question on hold still on topic? It doesn't really seem like a question any more, just a lot of grumbling.

Kudos to @MartijnWeterings and @NickCox for being far more patient than I'd ever be.

@El'endiaStarman, that should be OK. It may help people to make the context more concrete, or to provide a small example dataset to illustrate the issue.

@mkt, it continues to be a discussion about the site's policies & the status of a thread. So I'd say it remains on topic on Meta Stack Exchange. That said, a different question is whether the discussion remains productive. Note that no one is obliged to continue to engage in any thread beyond their level of patience.

@mkt Thanks for the support, much appreciated. In any discussion that gets awkward or even angry it can be hard to be sure who's being unreasonable, quite possibly everyone in most political situations. I have a constitutional weakness not to give the other person the last word if I think they are wrong. And another weakness to defend myself when attacked. I am really out of it now.

I think the OP would be better advised to spend the energy expounded in "whinging" (@mkt that's a shared word) on trying to analyse their data, which is the focus of the original question.

@gung Fair enough; I gave up on responding a while ago but should just ignore the whole thing.

@NickCox Agreed, I think it's far from productive at this point and you both have already gone above and beyond.

@gung Thanks for the feedback!