Description
As I recently learned in John Swinbank's review of PhotoCalib, the DPDD says this about "Err" and "Sigma":
For all catalog data, we will employ a convention where estimates of standard errors have the suffix Err, while the estimates of inherent widths of distribution (or functions in general) have the suffix Sigma
On the other hand, we use Sigma in many places where we really mean Err. For example, our source tables are littered with _fluxSigma and _xySigma, which are documented as "1-sigma uncertainty". Our database schema is mixed in this regard as well (e.g. raSigma, an "uncertainty" and uPSFluxSigma a "Standard deviation of the distribution" in the DiaSource table).
This RFC is to gather consensus about how best to sanitize our usage of these terms in the stack and related software and interfaces. The concrete propose here is to replace all incorrect uses of Sigma with Err, and vice versa in all code below lsst_distrib and in cat.
Attachments
Issue Links
- is triggering
-
DM-10933 Fix "Sigma" and "Err" in pipeline/developer documentation
- Done
-
- Done
-
- Done
-
DM-15244 Change fluxSigma to fluxErr and similarly for apCorr and covariances
- Done
- relates to
-
- Done
-
- Implemented
-
- Implemented
-
- Implemented
Activity
Implementation ticket filed as Epic DM-10932, with associated stories.
NOTE! When the schema and code tickets (DM-10935, DM-10936) are implemented, this will be a breaking change for backwards compatibility, including old reference catalogs and previously processed data. We recommend that publicly provided reference catalogs be copied and marked "deprecated: for use with LSST stack < v??", and that updated catalogs be provided.
The code has been fixed. Now I'm looking at our documentation. I think the only thing that needs changing is C++ style guide entry 3-34 says to use "Sigma" not "Err" for error. Could somebody please suggest new wording for this entry? I'm really not sure what to say...for instance do we still recommend "Var" as a suffix for variance?
Russell Owen: that was already fixed in DM-10933, it just needs to be merged (the agreement to merge happened just before the all hands and I missed it).
Lets keep discussion about remaining work on the relevant epic: DM-10932
Fully implemented by the completion of DM-10935.
Thanks for clarification, Tim Jenness and John Swinbank. I will work with Simon Krughoff tomorrow to file the appropriate tickets.