Few doubts in cognos

[sairaochowdary]

hi anyone
I have few doubts.any one could clear my doubts

1.what is meant by monster dimension
2.difference between factless and fact tables
3.what is meant by divide by zero and 80/20 suppression.
4.what is meant by associate level in trasformer

Thanks
Sairao chowdary

 

[drlex]

Sairao,
1. Not heard that before and sounds to be an unlikely Cognos-sourced term! Is it possibly just a user expression to indicate a dimension with a number of categories that run close to the maximum number allowed or a high number of levels?

2. that's one better answered in the Data warehousing General Discussion Forum Forum353. I'd say it was the difference between tables that contain transational data - sales and those that are more static - customer details, product details, supplier details.

3. a)divide by zero is a message indicating a mathematical error (whether the answer is infinity or aleph-null, I'll leave to mathematical logicians). In Cognos, Impromptu will return an error, so one needs to trap and exclude the possibility of a division by zero in calculated fields. In transformer, calculated measures that result in division by zero will display '/0' unless set otherwise.
b) 80/20 rule, so beloved of management, is a way of reducing the 'tail' of data so as to allow concentration on the major sources. I've understood it to mean that 80% of the total arises from 20% of the sources. However, in Powerplay, the algorithm acts to determine those rows or columns or both (as selected)that comprise 80% of the total and summarise the remaining 20% as 'Other'. You may find differences between PPWeb and PPclient until 7.3 is released.

4. Association in transformer is necessary to relate the datasource(s) to the model. Here is an extract from theStep by Step Transformer pdf:
'The process of defining the relationships between an item in a Transformer model (such as a level) and its associated columns, attributes, or tables in the data source. An association specifies a type (such as a column), and a role (such as a source, label, description, or drill
through target).'
If you have two data sources with the same item (eg Customer), you'll need to mark customer in the dimension as unique in order to allow transformer to associate the data from both sources appropriately.

lex

soi la, soi carre

 

[flex13]

1- Definitely NOT a Cognos term, so difficult to provide an answer unless another user is familiar with this term.

 

[john76]

I'll answer #2:

First let me explain what a normal fact table looks like, then a "factless" fact table should make sense.

A dimensional database model has (basically) two types of tables: Dimensions that contain categorical variables that you "slice & dice" your measures by, and Fact tables that contain the measures (such as Sales, Revenue, etc.) and keys that link these measures to the dimensions.

So, for example, you may have a fact table "Sales" with keys to dimensions "Stores", "Products", and "Dates". The fact table may have measures for "Amount", "Revenue", and "Profit". This is the standard fact table.

A FACTLESS fact table is one which does not contain any measures, but still links various dimensions together for some purpose.

HTH,
John

 

[monsky]

Factless fact tables are useful to describe events and coverage. The first type is one that describe events. A good example is a student attendance. The fact table contains the keys and perhaps one or more attributes that contain a constant of 1.

The second type is a coverage table. This fact table contains nothing but keys. It is useful in answering questions about things that didn't happen.

 

[johnherman]

Monster" dimension is sometimes used to describe a dimension with lots and lots of members. For instance, if you are a credit card company, it could be the account number or customer, which could be in the tens or hundreds of millions.

-------------------------
John Herman is available for short and long term data warehousing consulting and contracts.