Re: [OT] How to plot time series over weekly / monthly data
Hi George,
On Tue, Jul 28, 2020 at 07:35:11AM -0300, George N. White III wrote:
>
> Time-series plots just lead to questions about trends and seasonal patterns,
> so you are better off starting with a package that can decompose a series
> into seasonal patterns+trends+residual.
>
> For monthly data, R has ts and stl. Weekly data are problematic as weeks
> don't fit evenly into years. For ecological data I like 48 intervals per
> year, which
> gives 7 and 8 day intervals, but for Debian day-of-week might be important,
> so if you get serious you would look to the social sciences.
Thanks a lot for your extra considerations I'll keep in mind. However,
for the given purpose to demonstrate that the one week of COVID-19 sprint
in Debian Med team had a big effort its probably perfectly fine that way:
https://people.debian.org/~tille/talks/202008_online/debian-med/commits.pdf
The double peak is probably marking the migration from Alioth to Salsa.
I need to investigate the other peaks. Some kind of valeys (which can be
better seen in the monthly data
https://people.debian.org/~tille/talks/202008_online/debian-med/commits-month.pdf
are possibly correlated to freeze times. I'll mark Debian release dates
and Debian Med sprints which most probably explain some interesting peaks.
Just to let you know what I'm working on.
Kind regards
Andreas.
--
http://fam-tille.de
Reply to: