Re: Oracle Spatial Options and Data Warehouse

ces_at_xnet.com (Chuck Schuelke) writes:

>could someone use values such as treatment outcome instead of distance
>in spacial attributes.
 

>ie give me diagnoses and procedure codes where success = cancer in
>remission or better. or something like this to evaluate treatment
>efficacy?

Hi Chuck,

This confused me a lot when Oracle originally came out with their Spatial option. Back then (in 1995) they called it Oracle Multidimension and said in some press releases that it would eventually become Oracle's OLAP offering. My company which is in the OLAP arena had to sit down and figure out if Oracle really meant for Oracle Multidimension to compete in the OLAP marketplace or not.

After a lot of analysis, here's basically what we came up with to differentiate what Spatial vs. OLAP analysis is all about.

Spatial analysis is fundamentally used for asking questions like "Who am I near ?" or "What is next to me ? " The data itself is stored in continuous ranges - Such as altitude from 0 to 30,000 feet, pressure from 25 to 35 inches of mercury, Longitude and Latitude from 0 to 360 degrees, etc. When setting up a data element using HHCODE, the special encoding used by the spatial option, you define the range the data can have and the granularity of that range. All the possible values for a particular piece of data must fit within this range.

Multidimensional OLAP analyis is fundamentally used for asking very different questions against a different kind of data. For example "Show me the eastern market and all of it's descendants." or Show me my top 10 products by region in the first quarter of the year ranked by profits. The data is stored in discrete, not necessesarily linear, hierarchically grouped members in structures called dimensions. (Example the product Dimension might contain the following hierarchy of levels, each level made up of one or more members: All Products ---> Product Lines ---> Product Families ---> Product Groups ---> Products ---> SKU's.

When setting up a member, you define where it fits in the multidimensional model in relation to its parents and its children. OLAP data doesn't necessarily fit in a continuous range like spatial data does - for example there is no such thing as a continuous range of customers.

So in summary, if your problem mandates a need to slice and dice and drill up and down on lots of discrete values, ususally with heirarchical relationships and aggregated / calculated data you are most likely looking at an OLAP problem. If instead you need to examine relationships of continuous ranges of data based on nearness or location, you may be looking at a spatial problem.

I hope this helps.

-- 

- Dan

Home:                                           Office:
Daniel Druker                                   Daniel Druker
                                                Strategic Alliances
1427 Byron Street                               Arbor Software Corporation
Palo Alto, CA, 94301                            1325 Chesapeake Terrace
ddruker_at_netcom.com                              Sunnyvale, CA, 94089
                                                (408) 541-4027
                                                ddruker_at_arborsoft.com