[Webinar] Streamline your web hosting managementRegister Today

x
  • Status: Solved
  • Priority: Medium
  • Security: Public
  • Views: 110
  • Last Modified:

date of death fact tbl or dimension?

I need to report on date of death.  I think it goes into a fact table since it's something that has happened. But what if the date of death is null? How would I populate it. I don't want a null in fact tbl.   I really don't want to use a dummy date like 9999-12-31. Any suggestions?
0
elucero
Asked:
elucero
  • 7
  • 5
  • 2
  • +1
1 Solution
 
Jim HornMicrosoft SQL Server Developer, Architect, and AuthorCommented:
NULL would be appropriate here, as by definition it is the absence of a value.

>I think it goes into a fact table since it's something that has happened.
Not necessarily, as fact tables are measurements, of which date of death would be one.

Unless you want DimDeath, but then your Fact would still join on that table...
1
 
eluceroAuthor Commented:
Yes, ok NULL does make sense, but won't I have problem with NULL value in a cube?
0
 
Kent OlsenData Warehouse Architect / DBACommented:
It would depend almost entirely on your database design.

If it's a traditional OLTP, date_of_death would probably just be a column it the table that describes people.

In a data warehouse, the structure of the fact table(s) usually resembles the OLTP's transaction table(s).  Depending on the nature of the business the transactions may be split over multiple fact tables.  An insurance company would probably have 2 fact tables in their D/W.  One for policies/premiums and another for claims.  The two tables would have almost nothing in common except the policy ID so separate tables has advantages.

Does your D/W record multiple events for people?  Birth, graduation, marriage, etc.?  If so, a death record in the fact table is appropriate.
1
The new generation of project management tools

With monday.com’s project management tool, you can see what everyone on your team is working in a single glance. Its intuitive dashboards are customizable, so you can create systems that work for you.

 
eluceroAuthor Commented:
It's healthcare industry.  If death occurred based on specific disease.   I think it goes into fact tbl??
0
 
PortletPaulCommented:
the date of death is a fact

NULL (or no row) is a fact too (i.e. not dead, or no death recorded)

death as a dimension would not be very helpful as it is a boolean condition (is dead/is not dead)

a time dimension (I assume you have one) would be relevant to the date of death
(e.g. to answer "how many deaths in fiscal year 2014")
1
 
Kent OlsenData Warehouse Architect / DBACommented:
I can envision a lot of scenarios and the possibility of several fact tables.

How may fact tables are there?  How may of them record people specific events?
1
 
PortletPaulCommented:
I can envision a lot of scenarios and the possibility of several fact tables.
agreed!
1
 
eluceroAuthor Commented:
Thanks for the help.  I have one fact so far with case number, bunch of dates including dt of death which will mapped up to datedim.  I'm worried about putting the dt of death in fact because of how analysis services handles NULLS in a cube.  I don't want it to default to a dummy value like '9999-12-31'.
0
 
Kent OlsenData Warehouse Architect / DBACommented:
If the patient survived, don't put a row into the fact table that marks the date of death.  That would seem to solve the problem.
0
 
eluceroAuthor Commented:
I would have put the row in the fact if patient survived since other facts in fact table are caseID ,diaganosisdt, submitteddt, deathdt. So I still have the null issue.  Weather they died or survived I would still have values for the rest - caseID ,diaganosisdt, submitteddt.
0
 
Kent OlsenData Warehouse Architect / DBACommented:
Ah.  So death is a byproduct of another event, not an event unto itself.  (Never thought I'd hear death described that way, but what the heck....)

With that as a backdrop, the choices would seem to be:

1) Accept that an event that doesn't result in death will have a NULL in the date_of_death column.
2) Default the column to a specific non-null value.  (Very bad idea....)
3) Insert another row in the fact table to provide the death details.

So going back to the original problem, if there is a date_of_death column you're going to have to use 1 or 2 above.  It's not a classic star-schema design, but adding a dimension table on the date_of_death column solves your NULL issue as you'll only include rows with a non-null date in the dimension.  One of the byproduct will be that the optimal tests for survived/died will be slightly different as you can test the dimension table for died, but will have to test the fact table (or outer join the died dimension) to test for survived.  Personally, I'd include the NULL values in the dimension but that seems counter to your wants/needs.

Kent
0
 
eluceroAuthor Commented:
Thanks, so you're suggesting a death dim which would be like this
caseID
DateofDeath

Then I would do an outer join from factCases to deathdim for my schema fact.caseid = deathdim.caseid.  I think this will work than I can avoid the null values.  Why would add the null values in the deathdim? They only need to report deaths not survived.
0
 
eluceroAuthor Commented:
I'm thinking now that this would be a good junk dimension since I have other yes/no attributes like case closed, died, hospitalized?  So it would be like this

caseid
dateofdeath
death y/n
datehospitalized
hospitalized y/n
dateclosed
caseclosed y/n
0
 
Kent OlsenData Warehouse Architect / DBACommented:
If there's exactly on row in the fact table with any given caseid value, this can work.  If a case produce multiple rows any counts (or other aggregate function results) that are based on the dimension will be wrong.
1
 
eluceroAuthor Commented:
Yes, I think will work.  I'm going to go w/junk dimension.  Thanks for your help, Elizabeth
0

Featured Post

Hire Technology Freelancers with Gigs

Work with freelancers specializing in everything from database administration to programming, who have proven themselves as experts in their field. Hire the best, collaborate easily, pay securely, and get projects done right.

  • 7
  • 5
  • 2
  • +1
Tackle projects and never again get stuck behind a technical roadblock.
Join Now