r:ManagedMissingValuesRepresentation at RepresentedVariable level

[dzkwsk]

In DDI L 3.3 the r:ManagedMissingValuesRepresentation object can be used to add, at the value domain level of the instantiated variable, sentinel values such as: "out of scope", "doesn't say", "not concerned" etc. Would it be possible to add this object to the represented variable as well? Would it be possible to add this object to the represented variable as well?

[wlthomas]

Discussed at TC meeting 2024-04-25 After general comments for clarification the idea was raised that this issue should be viewed more broadly and initial discussion should be to clarify "What is the purpose of the RepresentedVariable?"

In the past there has been debate about what needed to go in the RepresentedVariable and what was added at the Instance Variable level. One approach is that If its available it should be in represented variable. This leaves specification of the Population within the Universe (addition of spatial and temporal specification) at the Instance level. Others that, capturing the categories of Missing Values made sense in the Represented but that coding should go in the Instance level as these can change significantly from capture to various storage structure (often collapsed or recoded over these processes).

Different sorts of data throw up different problems One of the strengths of the schema it is quite directional (this can also)

This discussion may also be affected by a possible change of the current approach of using Representation types for Response Domains. We should make sure and relay information from this discussion to the Question Group.

Members were asked to add comments regarding the role of the RepresentedVariable in Lifecycle (or in general).

[DanSmith]

The Represented Variable already allows defining missing values.

Code Lists can have their categories marked as missing. Numeric domains, datetime domains, text domains, etc can all already set missing value tokens.

This request is simply to additionally allow defining missing values using the managed missing value structure in addition to the currently allowed missing value definitions. Allowing managed missing values as the method for defining missing values seems totally reasonable, especially in light of the documentation for the missing value tokens, already present on Represented Variables in the values representations, that states Use of MissingValuesReference is preferred.