A common scenario is to have Enums that are bins over ranges of a scalar value
Each of the above has ranges encoded in the annotations but in different ways
Most range bin enums bin by the same unit, but this is not guaranteed, e.g.
E.g.
name: AgeGroupEnum
description: Standard age groups used in NIH clinical research, particularly NINDS
CDEs
from_schema: https://w3id.org/linkml/valuesets
rank: 1000
permissible_values:
NEONATE:
text: NEONATE
description: Birth to 28 days
meaning: NCIT:C16731
annotations:
max_age_days: 28
title: Newborn
INFANT:
text: INFANT
description: 29 days to less than 1 year
meaning: NCIT:C27956
annotations:
min_age_days: 29
max_age_years: 1We might even have different units in the same PV
Note these annotations are not necessary - we can use linkml-map to write conditional boolean rules for mapping. But this can lead to ugly boilerplate, and it is often better to keep the metadata about ranges in the PV metadata itself.
If this were standardized then we could automatically (a) infer bins from scalars (b) perform consistency checking when both are present.
In the above AgeRange case, we can imagine age_days or age_years? (linkml-map already has examples of how unit conversion between these can be automated.
We can imagine a few different ways to standardize this.
Nesting
INFANT:
text: INFANT
description: 29 days to less than 1 year
meaning: NCIT:C27956
annotations:
min:
age_days: 29
max:
age_years: 1Flat
INFANT:
text: INFANT
description: 29 days to less than 1 year
meaning: NCIT:C27956
annotations:
minimum_age_days: 29
minimum_age_years: 1My preference is for the latter. This would rely on linkml-map specifications for aggregate operations, which don't yet exist. But at least having a convention:
minimum_[scalar_slot] and maximum_[scalar_slot], where scalar_slot is defined using normal linkml mechanisms
Would help with human readability and consistency
A common scenario is to have Enums that are bins over ranges of a scalar value
Each of the above has ranges encoded in the annotations but in different ways
Most range bin enums bin by the same unit, but this is not guaranteed, e.g.
E.g.
We might even have different units in the same PV
Note these annotations are not necessary - we can use linkml-map to write conditional boolean rules for mapping. But this can lead to ugly boilerplate, and it is often better to keep the metadata about ranges in the PV metadata itself.
If this were standardized then we could automatically (a) infer bins from scalars (b) perform consistency checking when both are present.
In the above AgeRange case, we can imagine
age_daysorage_years? (linkml-map already has examples of how unit conversion between these can be automated.We can imagine a few different ways to standardize this.
Nesting
Flat
My preference is for the latter. This would rely on linkml-map specifications for aggregate operations, which don't yet exist. But at least having a convention:
minimum_[scalar_slot]andmaximum_[scalar_slot], wherescalar_slotis defined using normal linkml mechanismsWould help with human readability and consistency