Use the word 'group' when refering to a set of responses or observations

[pinkwah]

A response and observation refers to a single datapoint, while we are using these words to refer to groups of responses and observations (ie. POLY_RES/POLY_OBS are groups).

I propose we change the wording to response_group and observation_group in the database and REST endpoints, so as to disambiguate the terms.

[markusdregi]

I think the suggestion makes sense. If so, we should use the same term also for sampled values from distributions etc.

[pinkwah]

We had a tiny discussion afterwards where I reckonned the name "group" doesn't fit this particular concept. We're currently using the term to refer to grouping of observations in order to do some large-scale operations on them, or otherwise mark them somehow. While in the case of this issue, "group" would mean a set of observations/responses tied to a particular variable (eg. a particular sensor).

Not sure how to resolve it as none of the synonyms fit the variable definition as well as the word 'group', so I've left it hanging while browsing thesauruses.

Perhaps we could keep 'group' as the 'variable' definition, and use 'selection' for the thing @xjules is working on?

---

I think the suggestion makes sense. If so, we should use the same term also for sampled values from distributions etc.

As in parameters? They are a different problem in terms of naming. Each "group" of observations refer to a single variable that has been sampled multiple times. A "group" of parameters would refer to single samples of multiple variables, all related to each other by some relatively arbitrary physical property, eg. porosity. So the domain differs and thus would be perhaps a bit confusing to reuse the word 'group' (or w/e else we decide) for both.

If literature uses the same term for both things then I concede, since we should use the same terms as in the literature.

[markusdregi]

I think naming this groups and what @xjules is working on as a selection or perspective would be better :)

That would also be consistent (with respect to group) to the daily lingo of our users...

[markusdregi]

But parameters can also be grouped with respect to their layout and not their realisation and that was my point. The following could be considered a parameter group:

coefficients:
  a -> gauss
  b -> gauss
  c -> gauss

And the following could be considered a sampled group of the parameter group:

coefficients:
  a: 0.51
  b: -0.145
  c: 0:21

But I do understand that you would like to limit this discussion to responses and observations for now and that is perfectly fine 👍🏻

[xjules]

Comments from our meeting, which can serve as a source of inspiration regarding the naming conventions:

In ert2 response refers to a simulation result, represented as a matrix or vector of values, and not a singular value. At the same time, responses represent a group of different responses, ie. not a single response with multiple values. Therefore one suggestion would be to use response vector, response set or response group to refer when dealing with multiple values of the same response, while response selection (or even response vectors) would refer to a group of different responses put together by a configuration specs, algorithm or user interaction. In ert-storage this would project only to renaming the end-points at the moment. Nevertheless, this would change the user's point of view and would require more changes in ert and documentation I guess. So it's a reasonable question whether it makes sense in the first place. The same applies to observation term.

[pinkwah]

Delaying this due to lack of consensus.

[mortalisk]

I would think it makes sense to just know from context what we are talking about, so in an ensemble context, talking about a response, would be the entire ensemble of responses, while in a realization context a response would be only that realizations response of that same "group". What do you think @oysteoh ? Should we close this?

[oysteoh]

I'm not entirely sure , i think it stranded with the ert3 discussions and how to deal with this equally across ert2 and ert3. Do you have any information you can dig out from the back of your head @markusdregi ?

[markusdregi]

I suggest we resolve or close this issue as part equinor/ert#2707