[Bigbang-dev] activity in mailing lists and correlation to meetings, draft publications

[Sebastian Benthall]

>
> Just looking at those timelines qualitatively,

there are some connections to meetings,

where there’s more mailing list traffic leading up

to a meeting or following a meeting.
>

It's occurred to me while working on this task
<https://github.com/datactive/bigbang/pull/386> that we probably should
settle on a quantitative metric for this kind of correlation.

This turns out to be a quite complex issue.
There are many interesting ways to assess correlations in time series data.

None of them are particularly well suited to the *kind* of time series data
that we are using: where we have sparse binary events in continuous time.
I.e., most intervals have 0 events; there's a 1 in a single moment in
continuous time.
Using conventional methods on this kind of data requires a discretization
and aggregation
step which can introduce a lot of bias.

I expect that for the key audience for this work (Article 19), these kinds
of statistical details
are not of interest to them. (i could be wrong)

But if anybody would like to engage in more depth about the statistics of
time series
data and how we can accomodate these kinds of measurements in BigBang, I'd
be
very happy to work with them on this. It's a longstanding interest of mine.

- S
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ghserv.net/pipermail/bigbang-dev/attachments/20200626/174d4578/attachment.html>