- Discussion
- Statistics & Health
- Questions about UML model of Statistical Units

I have some questions regarding the UML model of Statistical Units.

I hope they are so trivial that someone knows the answers. :)

1. First of all, about Tessellation:

-Is it correct that in the model the AreaStatisticalUnit is same as any NUTS unit or other area unit, and that Tessellation is all the area units together (area units of one level)?

-In the model, the Tessellation is not mandatory for AreaStatisticalUnit (in topological sence/TopologicalConstraint).

-Does it mean that AreaStatisticalUnit can also exist alone?

-And if there are many AreaStatisticalUnits of one level, but there is not data available for all the units, some units can be missing from the data and therefore they don't form Tessellation?

-Or is it something else: For example in Finland we have coastline and then the statistical areas can be either polygons that cover the sea too (forming then tessellation) or they can be clipped with the coastline and be multipolygons (with islands). So in the latter case they would not form tessellation, since the sea areas between islands are lacking from the dataset?

-So all the grid cells have a grid to link, but why don’t all AreaStatisticalUnits have a Tessellation to link?

VS.

2. About VectorStatisticalUnits:

-Does VectorStatisticalUnit mean, that it can be any statistical unit, also line or point?

-But AreaStatisticalUnit has to be specifically NUTS or other area?

-And AreaStatisticalUnit is a subtype of VectorStatisticalUnit?

I appreciate, if someone knows the answers. Thank you!

BR, Tuuli

- Public

By Marcus BrühlDear Tuuli,

I had also asked for an explicit NUTS application Schema (see: https://themes.jrc.ec.europa.eu/discussion/view/180156/nuts-gml-application-schema). Unfortunately, there’s no attribute to distinguish different types of statistical units. For example, if you publish all NUTS regions as AreaStatisticalUnit it will be hard for users to distinguish NUTS1, NUTS2 and NUTS3. This was also addressed in another discussion topic: https://themes.jrc.ec.europa.eu/discussion/view/131523/missing-a-statistical-unit-type-in-the-datamodel

Another option would have been to select all NUTS regions of one level via the StatisticalTesselation. But the multiplicity of the association AreaStatisticalUnit – StatisticalTesselation is 0..1 (and not 0..*).

Concerning tessellation, it’s indeed not clear what’s meant with the voidability of the association. From modelling perspective, both of your interpretations seem to be possible. The intention of the TWG SU is not clear.

Concerning AreaStatisticalUnit, it’s indeed a subtype of VectorStatisticalUnit. But you can provide any statistical unit, which a surfacic reference geometry (in reality), not only NUTS regions. But, as far as I read the schema, you don’t have to provide the geometry as surface. The geometry itself is ‘hidden’ in dataType VectorStatisticalUnitGeometry. I could imagine to publish the NUTS regions as AreaStatisticalUnit (with an areaValue) with referenc points as geometry in dataType VectorStatisticalUnitGeometry.

In my opinion, the TWG SU decided to draft the schema very complex. I guess, modelling experts love it. But why to publish a rather simple theme like NUTS regions with such a complex schema?

Best regards,

Marcus Brühl / BKG Germany

- Public

By Tuuli PIHLAJAMAADear Marcus,

Thank you very much. That clarifies the thing for me a bit. So there is definitely a risk of different interpretations of the UML models.

I find it still a bit weird, that if the geometry is hidden inside VectorStatisticalUnitGeometry, how is there still TopologicalConstraint between AreaStatisticalUnit and StatisticalTessellation. But I'm not an expert on that, so maybe it makes sence and I will understand it later.

BR, Tuuli

- Public

By Bresters PIETERDear Tuuli and Marcus,

First of all, sorry for my late response.

Secondly, as a former member of the TWG SU-PD, I would like to apologize for the complexity of the model.

**About the Tessellation:**Your first assumption about the AreaStatisticalUnit being the same as any NUTS unit or other area unit, and that Tessellation is all the area units together, is right.

The AreaStatisticalUnit can exist alone without the Tessellation. In the Dutch harmonization of Statistical units, we misused the link to the Tessellation to solve another problem: The missing type, as Marcus also mentioned. This issue has been reported to the MIG, but the status is unknown to me. For the rest, we did not use the tessellation, because we don't have it as a feature.

About your last assumptions on the tessellation, it is right that a tessellation can’t have gaps and can't have area statistical unit that overlap according to the definition.

I don't have an answer to the question "why all the grid cells have a grid to link, but why don’t all AreaStatisticalUnits have a Tessellation to link?" It seems inconsistent. In the Geostat 3 project we did not use the feature type grid, only the grid cells.

**About VectorStatisticalUnits:**Below I give answers to your last 3 questions in red.

-Does VectorStatisticalUnit mean, that it can be any statistical unit, also line or point? YES

-But AreaStatisticalUnit has to be specifically NUTS or other area? I suppose it has to be an Area because it is in the name, but also to let the extra attributes make sense.

-And AreaStatisticalUnit is a subtype of VectorStatisticalUnit? Yes, a subtype with some extra attributes like given in the model

Best regards,

Pieter Bresters,

Statistics Netherlands

Join this group if you would like to share knowledge or ask questions regarding the INSPIRE implementation of Statistical Units [SU], Population Distribution (Demography) [PD] or Human Health and Safety [HH] data themes.