(Feedback from Statistics Sweden; 2 October, 2017)

The Business part is abstract and can be more detailed for the objects Statistical Support Program, Business Process, Statistical Program, Statistical program cycle. These objects need to handle new ways of producing statistics, like one Statistical Program handling collection and microdata editing and several other Statistical programs using the result.



  • No labels

3 Comments

  1. user-8e470

    ModernStats Workshop 12/4: Agree this deserves review. Several organisations are implementing this part of model and have made additions to the model. These should be compared to see what should be added to GSIM. Francine Kalonji Eva Holm Alistair Hamilton

  2. Alistair Hamilton

    Hi user-8e470.  The notification came in just before I went on leave and I am only now wrapping up the "non burning" emails that were waiting when I got back.

    Speaking just from ABS experience, while we have done lots of implementation work in this space I am not sure how much of it bubbles up to additional conceptual modelling in GSIM. 

    For example, pressing questions for us are of the type

    • when do we have one Statistical Program (SP) vs two related SPs?
    • in terms of SP Cycles, if we do one set of things monthly, some special things quarterly and some super special things annually - but it is all in aid of the same monthly statistical output - do we just have a single set of monthly SPCs?

    While I'd be super interested in how other implementers have approached such questions I am not sure GSIM would add more value by trying to develop additional "universal" modelling for an aspect that possibly varies more from NSO to NSO depending on their business model, scope, preference etc than any of the other three quadrants of GSIM?

    You'd have seen from Justin this applies to the "process" bit of the GSIM Business quadrant as well at ABS, shaped by our investment in metadata actively driving a BPMS - which will not be a "given" across all implementers of GSIM.  As Justin highlighted, this led us to do a lot more modelling around - in GSIM terms - Process Input and Process Output but the driver was our "transformation (DX?) architecture" rather than conceptual.

    The "DX" driver also led (eg) to lots of additional logical modelling of connections between SPs/SPCs and

    • acquisition events,
    • data structures and data sets of specific types etc.

    Once again, this was done in the context of how ABS wanted to operate, with some weight even being given to whether a particular approach aligned well or poorly with key COTS platforms supporting our DX.  It <might> be possible to abstract something more universal and conceptual from the logical modelling but I am not 100% sure.

    What are other implementers thoughts?

  3. Guillaume Duffes

    Hi,

     

    We indeed have an extension of GSIM in this area. I remember discussing that point with Francine during the last workshop in Geneva. Our main additional classes deal with what we call the statistical operations (statistical activities in StatCan). You can below the additional class in beige (the other ones are the original GSIM ones), sorry it's in French, the English version has not been updated for a while now...

    I remember also a very rich conversation with Alistair, Jay, Flavio and others by e-mail on a very close subject. I have just forwarded it to you Thérèse and Inkyung. It might go a little bit too far for this issue, I don't know.

     

    Regarding the original question, it's hard to say whether or not a deeper conceptual model could be adopted in this area. However, what I am pretty sure it is huge piece of work, and can't be handled in the short time left (smile)