Snowflaking the Date Dimension

I understand the OLAP sayings of "make the facts narrow and the dimensions wide" and "denormalize", etc. But I am a novice. I inserted a graphic of the table structure, but I'm not sure how to go through that process when
it involves a date dimension. Here's a little background:
A facility has a name, and multiple units
A unit has a name and a facility
For each date, a facility has the number of patients residing
For each date, a unit has the number of patients residing
A drug has a name
A drugroute as a name and a classification
Each fact is an instance of a drug used on a specific date and in a specific unit.
The number of patients residing in a facility on a day does not equal the total number of unit patients, because patients are only counted once at the facility, even if they are listed in multiple units in the same day.
A drug dispensed multiple times on the same day in the same unit only counts once, unless stratified by route. When route is reported, each route administered is counted.
The most important question is: how do I NOT snowflake the date dimension?
Thanks, Kevin

So what do you think a snowflake schema is? Because your question does not make a lot of sense.
In Data Warehousing a snowflake schema is one where a dimension is normalized.
So in a star schema you might have a geography table with 4 columns
geographyID
Country
State
City
Where as in a snowflake schema you could have 3 tables, one each for Country, State and City.
So in your example you would have one date dimension (which could have columns for Year, Quarter, Month etc) which would have a grain at the day level. And it would be linked to each of your Facility, Unit and DrugDispensed fact tables. There is no need
to snowflake date here. It is perfectly fine for a dimension to be related to more than one fact. (in fact you should aim to use shared dimensions wherever possible)
http://darren.gosbell.com - please mark correct answers

Similar Messages

Date Dimension not updating correctly

Hi
My current date dimension underlying data was due to expire soon so I set about trying to create a new one.
I got hold of the script that it was based and went about re-creating the data like for like.
I pointed the date dimension at this new data structure and it all processed fine.
When I was doing some testing I noticed althought the total figures were fine, some of the year figures were a little off compared to the underlying data.
I found that dates such as 31/12/2012 were being classified as year 2013. The month was also being classified as January but the date was still showing as 31 December 2012
I switched the date dimension back to the old structure and all year numbers were fine.
I decided to recreate the Date Dimension in the cube using the same structure as the current date dimension and when it processed with the new data structure all the numbers are correct.
I tried clearing cache but I am baffled as to why the new data structure won't work with the existing Date Dimension but will work with a newly created date dimension in the same cube?
Regards

Hi
I seemed to have solved it (although I am still testing), there was a data diffrence in the week key and that it was over lapping with the end of the year so I changed it to match what I already had.
It still doesn't explain that before making the change when I created a new dimension it seemed to work ok.
I had a look at the XML and the only difference that I could see was that the original date dimension had a Time Binding that was generated on the server or the data source.

Date Dimension in Oracle CRM On Demand Opportunity History Reporting

Hi All,
I have a small quesiton concerning the Date Dimension in Oracle CRM On Demand's Opportunity History analysis.
To which date field does this dimension/column relate? Created Date?, Modified Date?, etc
Thanks in advance

Pl post this in a Siebel related forum
HTH
Srini

Problem populating date dimension with correct values

Hi all,
I have a 'simple' problem, but can't get it solved! The YTD calculations in the cube don't 'reset' at the right moment in time (they should 'reset' on the FISCAL year, but it actually seems to happen on the normal calendar year).
Our dimensions and cubes are MOLAP. There is a relational table, containing columns with values for both a CALendar hierarchy as well as for a FISCAL year hierarchy. This table populates the date dimension (only the fiscal hierarchy as in MOLAP we only have a fiscal hierarchy). The result looks correct in the data viewer.
The cube is populated from relational sources as well. One view provides the measure values and the CODE (business identifiers) values for each dimension involved. In case of the date dimension, it provides the code value for the involved level at which we load the cube (month).
I made a very simple testcase with only 1 dimension (date) and 1 small cube. The cube uses only the date dimesion (to load cube on the fiscal hierarchy at month level) and has 1 base measure that is loaded with a simple number and 1 calculated measure that calculates the YTD value for this base measure. The calculated measure is added in the OWB cube designer using button "Generate Calculated Measures" and choosing the "Year to Date" function.
When populating the cube and using the data viewer to check the results, the YTD measure values don't 'reset' at the fiscal year end. They seem to reset at the normal CALendar year end!?
After some testing, I have concluded it has to do with the values I supply to populate the date dimension, but I can't figure out what the modification should be and I can't find any examples anywhere.
Does anyone out there have a working YTD calculation in MOLAP?
Any help much appreciated.
Regards,
Ed

Sorry David, I am not sure I've got this right yet. I am looking at using the cumsum function, but I am not sure how to apply this. That is, I can see examples (cumsum(measure, total_dim, reset_dim)), but I think I cannot put such formula directly into the expression editor for a cube measure in OWB?
Until now I have been using the standard available calculated measures in the OWB measure editor and I think they insert an xml template? I'll get the Excel Calculation Tool from OTN, to see if I can create the correct formula and paste that in OWB in the expression editor for the measure. (Does that tool need a ROLAP workspace or does it need / can it work with a MOLAP workspace?)
Thanks and regards,
Ed
Edited by: EdSp on Jan 13, 2009 1:57 PM
I have downloaded the Excel sheet and can see how it works, but it seems to "insert" the calculated measures straight into the AW. However we deploy the AW from OWB so I will be looking at getting something to enter/paste in the expression editor of OWB...

Querying last element in a dimension based on another level in the same dimension using MDX

Summary: I need to write an MDX that returns the last date available in the level in the date dimension based for each month in the same dimension.
Details: i have the a dimension which has the following levels.
Dim_Date : Date (key) , Month (name of month), Quarter, Year.
The query i m trying to come up with would return the result something like
Month || Max_date || measure1
Jan 2014-01-25 100
Feb 2014-02-28 200
Please let me know if the above makes sense. Basically i m trying to get the max date for each month.
Thanks
Sri

Hi SrikanthGS,
Assuming that with the last date available for the measure you mean the last date with non-empty figures, consider the following against the AW:
SELECT [Measures].[Internet Sales Amount] ON 0,
Generate (
[Date].[Calendar].[Month].Members,
( Exists ( [Date].[Month of Year].[Month of Year], [Date].[Calendar].CurrentMember ),
Tail (
NonEmpty (
Descendants ( [Date].[Calendar].CurrentMember, [Date].[Calendar].[Date] ),
[Measures].[Internet Sales Amount]
1
) ON 1
FROM [Adventure Works]
WHERE [Product].[Product Categories].[Subcategory].&[1]
Philip,

SSAS Date Dimension attribute not displaying all members when browsed singly

I have an SSAS Cube database that holds data from 2009 to 2013 (5 years)
There is a Date Dimension in the database which holds a "Calender Period" Hierarchy with the attributes CalenderYear >> Quarter >> MonthYear >> Day .
I am connecting to the cube via Microsot EXCEL
Whenever i try to browse this hierarchy alone in the spread sheet, the first attribute as per the hierarchy shown above i.e. Calender Year does not display all the members, that is confusing the users to assume that the cube is not processed with the latest
data, which is not the case. However as soon as i try to browse any measure with this dimension, the data automatically shows for the latest dates of 2013 as per the expectations.
The perspective that is selected while connecting to the Cube in this case is the one that is by default the one including all the Cube Objects.
Whenever i try to browse the same cube for the same Dimension Hierarchy but in a separate user perspective i get the latest three Calender Years i.e. 2011,2012 and 2013 but not the complete years 2009 to 2013.
When i was browsing this earlier by the original base perspective (which has all the cube objects visible) the Calender Years displayed were 2009,2010,2011&2012 but not 2013 which should be there.
I checked the base view in the DB that is used to create this Dimension, that is fetching the complete data for 2009 to 2013.
I have also verified that this issue is still there if i try to browse the cube from the SSMS.
Can anyone help me to get all the years from 2009 to 2013 (the years for which my cube has data) become visible as soon as i drop it alone into the Spreadsheet?
Thanks
Suvrat

Thanks for your response Muthukumaran.
I did not know about this default measure property of the SSAS cube, it would be nice if you could also tell me how to check which is the default measure for a cube.
I understand your point that if a particular measure group does not have data for a year then that particular year will be skipped as per the default excel settings, which when disabled allowed all the years to be displayed, but in that case every time the
users connect to the cube via excel, they will have to configure this property.
I would like to share the way i resolved this issue:
I went ahead and simply did a Process Full for the Date dimension. The years started to display by default as per the expectations. Now after observing this change, as per my understanding I believe this issue was due to the CACHE of the cube not being updated
for the date dimension. As after each incremental load of the DWH we processed the cube & while processing the cube the dimensions were set for a Process Update and the Measure Groups were set for a Process Full. This caused this issue, and we could resolve
this once the Date Dimension was processed full.
Thanks
Suvrat Rai

Naming of multiple Date Dimensions - Logical Layer

I would like to get some advice on the following:
I am currently in the process of organizing all the dimensions and facts. During the organizing I have 12 different Date Dimensions. Some facts are join to 6 of those Date Dimensions. With each of the date dimensions, all of them have the same number of columns and names. Therefore, if I am creating a report, carry over the month of Delivery Date Dimension - Calendar Month lets say and Cancel Date Dimension - Calendar Month and then proceed to the results tabs, you will see two columns "Calendar Month" You wont know which one is from the Cancel Dimension or the Delivery Date Dimension. So we have to manually rename them in OBIEE. But I would like to rename them in the BMM/logical layer instead so the end user doesn't have to rename them each time. What I would like to get advice on is how did you implement the naming of multiple Date Dimensions. Cancel Date Calendar Month or Cancel MM etc? I don't want it too long then the column is very wide in the report.
Any advice is much appreciated!
thank you,
Jonathan

Thank you,
I read almost every best practice, but not much information around the Date Dimensions, just wondering if anyone has some great ideas around the Data Dimensions?
Thanks
Jonathan

Single date dimension while creating Aggregate tables

hi Guys ,
I have single date dimension(D1-D) with key as date_id and granularity is day-level. I have fact table(F1-D) which gives daily transactions . Now i created three aggregate tables with F2-M(aggregated to monthly),F3-Q(Aggregated to quarterly),F4-Y(Aggregated to yearly). As I said.I have a single date dimension table with date-id as a key.I have other columns month,quarter,year in Date dimension.
My question is : Is this single dimension table is enough to create the joins and maintain the BMM layer .I have joined the date_id to all the facts in physical layer. In BMM Layer i have one fact and 4 logical table sources .Ii have created dimension hierarchy on Date dimension and created the logical levels as year ,quarter,month and daily level and also set their respective level keys . Now after doing these i have also set the logical levels to the 4 logical table sources in the fact table.
Here i am getting an error saying :
WARNINGS:
BUSINESS MODEL Financial Model:
[39059] Logical dimension table D04_DIM_DATE has a source D04_DIM_DATE at level D04_DIM_DATE Detail that joins to a higher level fact source F02_FACT_GL_DLY_TRAN_BAL.F03_FACT_GL_PERIOD_TRAN_BAL
Can anyone tell me why am i getting this error .

1) as you suggest, creating mini dimension tables for each level
3) using views with select distinct x,y,z to model 1) but without the need for phsical tables
I have thought of these scenarios .But it's just creating a multiple Date Dimension tables with there respective grains if we follow 1 and 3 steps you mentioned.
But i don't want to create aliases or views either on Datawarehouse or physical layer (Admin tool).Just with using a single Date Dimension table I need to comeup with a solution as mentioned by my Colleagues.
2) pushing all the calendar attributes into the aggreate table (eg your year aggregate has all the year fields, your month aggregate has all the year fields and all the month fields etc.). Then in BM&M use the aggregate table as a logical table source for the date logical table.
What do you mean by "pushing all the calendar attributes into the aggreate table".If possible can you eloborate a bit so i can try in this way to get this done .
One morething .Here What they were saying is .Join the Single Date Dimension to all the 4 fact tables in Physical layer . Then drag all the tables into the BMM layer and create a Dimensional hierarchies for the Date dimension with levels as Year -->Quarter--> Month--> day level. With the Fact tables: Single fact tables with multiple logical table sources .Then we need to set the levels on the respective logical table sources with the desired levels .Say monthly aggregate table--> Month, Quarter Aggregate table--> quarter level in the Contents TAB.
When i do this I am getting the Warning:
WARNINGS:
BUSINESS MODEL Financial Model:
[39059] Logical dimension table D04_DIM_DATE has a source D04_DIM_DATE at level D04_DIM_DATE Detail that joins to a higher level fact source F02_FACT_GL_DLY_TRAN_BAL.F03_FACT_GL_PERIOD_TRAN_BAL

Copy/pasting date dimension to different project

Hi all,
We need to copy the date dimension from one project to another. Normally copy/paste can be used, but... not on the date dimension! I tried export/import with mdl but that imports it back to the same project?!
Thanks,
Ed

Hi Ed,
look at this thread (steps for importing MDL to different project name)
Re: How to Import Metadata into another projec
Regards,
Oleg

Project Portal, MSP_Portfolio_Analyzer, date dimension faulty week number

Hi,
I am working on a MDX query on the MSP_Portfolio_Analyzer cube of Project and I have problem that the date dimension contain wrong week number. For instance if I browse the dimension to 2013, quarter 3, month September it says day 1 is on Week 36. When I
use datepart function to return week number for 2013-09-01 I receive week 35 which is supposed to be correct.
The format on one date from dimension is:
[Time].[Time].[Year].&[2013].&[3].&[9].&[36].&[2013-09-01T00:00:00]
The MDX query I am building is going to show forcast for capacity and worked hours, but I have problem when the week number is off compared with week number retrieved from now(). Any suggestion on how to solve this?
[Time].[Time].[Year].&[2013].&[3].&[9].&[36].&[2013-09-01T00:00:00]

Hello,
This might help:
http://blogs.msdn.com/b/brismith/archive/2010/02/26/addressing-iso-8601-week-numbering-in-project-server-2007-and-2010.aspx
Paul
Paul Mather | Twitter |
http://pwmather.wordpress.com | CPS

Unable to filter the data for multiple time selections by dimensions

Hi to all,
I am new in MDX, i have a problem with my MDX query.
Calculated Member Logic:
SUM((OPENINGPERIOD([Date].[YQMD].[Year],[Date].[YQMD].[Month].&[2010-12-01T00:00:00]):[Date].[YQMD].Currentmember),[Measures].[Paid Amt])
Mdx Logic EX:
With Member [MEASURES].[Received_Amount]
AS
SUM((OPENINGPERIOD([Date].[YQMD].[Year],[Date].[YQMD].[Month].&[2010-12-01T00:00:00]):[Date].[YQMD].Currentmember)
,[Measures].[Paid Amt])
SELECT {[MEASURES].[Received_Amount]} On Columns
,[Date].[YQMD].[Year].members On Rows
From [Financial]
If i select multiple time periods in Rows, the query working fine.
but if select multiple periods in where clause it is not responding.
With Member [MEASURES].[Received_Amount]
AS
SUM((OPENINGPERIOD([Date].[YQMD].[Year],[Date].[YQMD].[Month].&[2010-12-01T00:00:00]):[Date].[YQMD].Currentmember)
,[Measures].[Paid Amt])
SELECT {[MEASURES].[Received_Amount]} On Columns
,[Speciality].[Specialty Name].[Specialty Name].members On Rows
From [Financial]
Where {[Date].[YQMD].[Year].&[2012-01-01T00:00:00],[Date].[YQMD].[Year].&[2013-01-01T00:00:00]}
Note:
Each of them is considered from the minimum date in the database to the selected time.
And also the data has to be filtered with respect to each drill down dimension.
If select multiple time periods the same formula has to be applied with respect to the dimensions.
Kindly help me to get out of this problem
Best Regards,
Nagendra

Hi David,
Thanks for your response.
I have a measure, i have to get received_amount in the database from the database starting period to my selection period. Later i have to check by dimensions using same measure. if i select any one period by dimension it's coming, but if i select multiple
periods in filter level by dimensions it's showing no records.
For Ex:
I have four years data in my database (2010-2013).
In 2010
Bill_Amt
Bill_Date
Specialty received_amount
1000
10/01/2010 4
600
2000
04/08/2010 2
1000
In 2013
Bill_Amt
Bill_Date Specialty
received_amount
1500
22/02/2013 2
1200
2000
14/03/2013 1
800
In the above scenario,
By Period:
if i go by period i should get,
Jan'13 ---> 1600
Feb'13 ---> 2800
Mar'13 ---> 3600
Specialty By Single Period:
If i select Jan'13 by specialty
Specialty
received_amount
2 1000
4 600
If i select Feb'13 by specialty
Specialty
received_amount
2 2200
4 600
If i select Mar'13 by specialty
Specialty
received_amount
1 800
2 2200
4 600
Specialty By Multiple selection Periods:
The result should be sum of the individual selection periods by specialty as follows,
If i select Jan'13 & Feb'13 by specialty
Specialty
received_amount
2 3200
4 1200
If i select Jan'13 , Feb'13 & Mar'13 by specialty
Specialty
received_amount
1 800
2 5400
4 1800
Regards,
Nagendra

Percentage based on the Date and Dimension - WEBI

Hello,
I have to calculate percentage based on the Dates and a Dimension.
I have BEX Query as source and reporting is done on BOXI - WEBI
Here is the table structure in the report.
List# , Process01, Process02.
List# is unique and Process01-Has 4 operations, Process02 has 3 operations (and these Operations can change from 3 to 7 or 2).
Operations will have a 'Date' only when it is completed.
and when all the operations in a 'Process' have 'Date' then i need to get the 'Latest Date'. If few of the operations are not confirmed, then i need to get the percentage like when 2 of 4 operations have date it will be 50% and when no operation is confirmed then 0% and so on .
Example Cube data:
List#,ProcessCategory,Operation,Date
100,Process01,456,11/11/2010.
100,Process01,457,11/12/2010.
100,Process01,489,
100,Process01,490,
100,Process02,890,11/13/2010.
100,Process02,891,11/14/2010.
100,Proceess02,892,11/15/2010
Result:
List#, Process01, Process02
100,50%,11/15/2010.
Please help me in this, i tryed different logics but it is giving ERROR at percentage.
Your help would be appreciated!
Thanks in advance.

I applied this:
If([Process Category] InList("Process01") And IsNull([Date]);[Percentage];Max(Date)))
[Percentage] = (Count([Date])/Count([Operation]))*100.
but, this is not becoming true - even when Operation has one Date
Showing result as:
List#,Process01,Process02
100, 11/12/2010, 11/15/2010
But result should be
100,50%,11/15/2010
Can anyone help me achieve the right result!
Thanks in advance.

SQL for adding data linked to the filtering dimension

This is from the Recipient RFM example where all Recipients are pulled in the initial Query and then we add data linked to the filtering dimension to compute recency, frequency and monetary.
INSERT INTO wkf505995918_2_2 (dPurchaseHistoryMonetary,tsPurchaseHistoryRecency,iPurchaseHistoryFrequency,iPurchaseHis toryKeysPKey0) SELECT DISTINCT Sum(P0.dAmount), Max(P0.tsDate), Count(P0.iPurchaseHistoryId), P0.iRecipientId FROM NeoPurchaseHistory P0 JOIN wkf505995918_2_1 W ON (W.iLinksKey1_0=P0.iRecipientId) WHERE (P0.tsDate >= add_months(CAST((GetDate()) AT TIME ZONE 'America/New_York' AS date), -(18))) AND ((P0.iPurchaseHistoryId > 0 OR P0.iPurchaseHistoryId < 0)) GROUP BY P0.iRecipientId
Note the JOIN between the linked NeoPurchaseHistory table and the wkf505995918_2_1 table populated by the initial query. The JOIN is ON the Recipient ID.

it doesnt matter what you name the file. oracle will be able to use it. if you need to change the name to meet your naming conventions, then you will need to take the db down and rename the file at the OS level and then startup mount to rename the file at the DB level.
why are you puting files in /dev? Thats for devices.

How to add Dimensions to the data in OEID

Hi ,
I am relatively new to OEID with some background on Endeca web commerce.
We are trying to import the data from OBIEE server into the OEID . We are able to achieve this by using Load Data from OBI server option in Integrator.
The issue we are facing is that all the attributes are under navigation pane in studio when we load the data source in a studio application. What we need to have is to remove some of the attributes (that are not managed) and add some managed attributes (a.k.a dimension).I understand PDR and DDR need to be worked on . But my question is that how is the association of attribute to the PDR and DDR managed?
Any help would be appreciated. Or any pointer to the document that can help in this will also be great as I could not find documentation to add dimensions(managed attributes)
Version : OEID 3.0
Regards,
Mayank Batra

Mayank,
Is it mandatory that I have OUTER_TRANSACTION_ID (even if I try to run this current graph only) ? The OUTER_TRANSACTION_ID parameter is necessary if you want to run the graph as part of an outer transaction. Configuration graphs such as this one are often included in outer transactions used to refresh the data domain. You can omit it from the Web Service Client during development, but if you decide you want to include the graph in an outer transaction, you will need to add it.
I am adding a Web Services Client and linked to the input port of this WSC ,the output port of the transformer. Where does the code mentioned in the link goes?Do we need to keep the code in a file and pass the URL to the WSC because looks like that is what it takes. As stated earlier in the topic, graphs used to load the standard attribute schema generally follow a standard pattern. See http://docs.oracle.com/cd/E39272_01/integrator.300/integrator_users/src/ci_csv_based_configuration_graphs.html for details about these graphs.
You do not need to store the output of the Transformer as a file. The XML document output by the Transformer component is held in memory during processing.
Note, BTW, that your graph should also include a Denormalizer component between the Transformer and the Web Services Client.
Basically, once the input is read into the graph, it is processed by a Transformer component to create an XML document that can be incorporated into a Web Services request.
The XML document output from the Transformer component includes line breaks. The line breaks cause the input Web Services request to be terminated incorrectly. Therefore, the output from the Transformer should be fed into a Denormalizer component and converted into a single XML string. This string is then added to the Web services request in the Web Services client and submitted to the Endeca Server.
RLJII

Hide dimension in data form also hide the data

Hi all,
It might sound simple, but i still could not found a workaround for this problem.
so here is the problem:
1. example i have dimension lineItem with member line1, line2, line3
FY13
Jan | Feb | ...
line1 | 1 | 1
line2 | 2 | 2
line3 | 3 | 3
2. i want to hide the line information so i choose to hide dimension for dimension LineItem. but instead hiding the dimension, it hide all the row (so no data shown).
3. i try to hide dimension that is used as column but everything is fine as the following:
FY13
line1 | 1 | 1
line2 | 2 | 2
line3 | 3 | 3
Anyone have this experience or method about how to hide dimension in row but still keep the data?
(i'm using hyperion 11.1.2.2)
Best Regards,
Feri

Hi SM,
Thank you for your help, actually i have another dimension in row as the following:
FY13
Jan | Feb | ...
new sales | product A l line1 | 1 | 1
l line2 | 2 | 2
l line3 | 3 | 3
i already test it again and i get the following result:
1. if i only have 2 dimension in row: product and LineItem (hidden) i get the following:
FY13
Jan | Feb | ...
product A | 1 | 1
product A | 2 | 2
product A | 3 | 3
(it work fine)
2. if i have 3 or more dimensions in row, i get the following:
FY13
Jan | Feb | ...
new sales | product A | 1 | 1
(it only show one row which is the line1, it seems like it cannot handle if we put more dimensions)
Edited by: Feri on Feb 27, 2013 3:12 AM

Snowflaking the Date Dimension

Similar Messages

Maybe you are looking for