Help with understanding of SAS VA LASR Star Schema?

[kjohnsonm]

Hello all,

I have some data in a fact tables, the keys of the data is stored in our Data Warehouse typically in three or more separate fields, I am trying to bring it in to SAS VA, (done) then link it up to some crosswalk descriptive tables. So I have been reading and trying to use this LASR StarSchema process. I have looked it up and followed the example but it seems this StarSchema will let me link more than one pair of tables at one time, but not more than one key per "table pair". Thus I went and looked for defining keys in SAS VA. I can see how to build indexes on more than one field in SAS Management Console, Data library manager, on a local library, for a single table properties, but the key tab I cannot get. (just escapes me, nothing active to click on/ or field to type in after selecting table)  (but I am not sure the star process would use this anyway)  Anyhow can this direction work or do I have to make a concatenated* key before importing?  or are their other ways to get at what I want that I have yet to see? I am willing to read, just not sure what to read next.  lol   TIA  -KJ

*or surrogate key

[alexal]

@kjohnsonm,

 but not more than one key per "table pair"

That's correct. The relation between fact table and dimension table is expressed by one pair of keys-that is, one variable in the fact table defines the relation to one variable in the dimension table.

[kjohnsonm]

Alexal,

So the Key field in the area described has no relationship here? and no other option save making a surrogate or composite key external to SAS VA?

If true; is there any chance you know where this is documented so I can show my supervisor?  PS thanks for your time.

[LinusH]

Just for mine (and perhaps others) understanding: are you referring to role playing dimension?
Typically if you have several date (keys) in your fact, connected to a generic time dimension.
This is a common design pattern and if VA haven't support for it, IMO it should be high up on the wish list...

[kjohnsonm]

Full disclosure, I am not 100% sure what "role playing dimension" is in this case related to my data. I am very new to VA, and somewhat new to Data warehouse world, I am a long time ERP transactional guy. My data is higher ED related, snapshot data regarding Admissions, I am working with term, snapdate, personID, and sometimes admit term all as keys in separate fields, then 70 or so facts that change from week to week as they (the applicants) process through the admissions process. We are using this data to help predict FTE year over year.  Sometimes the data per table might have more keys or less. All fact tables with personID will always have the snapdate for time comparison. We want to be able to look at where all our applicants’ counts on any given date relative to this coming fall term to last years and the year before etc. It’s to help admissions, budgeting, recruiting management, and administration

[SASKiwi]

As well as exploring star schemas as an option in VA, I strongly suggest you also take a look at just creating separate tables for your different dimensions and trying those out in some VA reports. VA reports can use multiple data sources and you may find you can satisfy your requirements without getting into the complexity of star schemas. You need to be careful of the performance overhead of star schemas as well - it can be tricky to get them to perform well.