Dear list,
I have 10-minute statistics (min, max, mean, stdev) data for which I know
the sampling frequency and therefore the number of samples.
I also (sometimes) have per-variable absolute and relative uncertainties
(JCGM terminology).
How do I encode the sampling frequency and number of samples in a way that
these are accessible in a standard way?
I now use, e.g.,
variable.cell_methods = "time: mean (interval: 10 minutes comment: sampled
at 1 Hz, so 600 samples)"
but this puts the frequency information in an ?opaque? comment. This makes
it hard to use that information for automated processing (an example of
this will appear below).
I'd like to do something like
variable.number_of_observations = 600
but that is not covered by the conventions, it seems.
Furthermore, because I have accuracy information (for the original
samples!) available, I'd like to add that in a standard way as well,
something like:
variable.uncertainty_absolute = 0.1
variable.uncertainty_relative = 0.01 # 1%
Then for each of the statistics a per-value uncertainty (estimates) can be
calculated. Let x be the value, a the absolute and r the relative
uncertainty, and n the number of samples, then for example
for min: a_x = a + r?x
for max: a_x = a + r?x
for mean: a_x = (a + r?x) / ?n
for stdev: a_x = (a + r?x) / ?n
I have seen in the archives that conventions for uncertainty information
have been proposed, but they seemed to refer to ancillary variables
(extensions of the standard_error standard name modifier?) or descriptions
of a statistical model. Both appear overkill for my purposes.
Best regards,
Erik Quaeghebeur
P.S.: I tried to unsubscribe as instructed from the
cf-metadata at lists.llnl.gov list that forwards trac discussions, but got a
Delivery Status Notification (Failure) message:
> The following message to <majordomo at lists.llnl.gov> was undeliverable.
> The reason for the problem:
> 5.1.0 - Unknown address error 550-'No such list.'
--
https://ac.erikquaeghebeur.name
Received on Fri Mar 16 2018 - 05:27:01 GMT
