Help me understand the Data Point object

Hi Petros, I don't know the answer but I will still add my comments here in case others (beginners like me) reading this post later have a similar way of thinking (and also understanding non-answers help understanding answers )

One question about your example, aren't "Gender {M, F}" and "Age {-20,20+}" Measure Components? Or is combination of these two Measure Components are being used as an Identifier Component?

Anyway, whether they are Measure Components or Identifier Components, sheet area A2:D5 contains values of Instance Variables, so it seems for me both white and blue cells are Data Points.

And regarding the cardinality, Measure Component is role given to Represented Variable, and at least for Unit Data Point, explanatory text says "This placeholder may point to multiple values representing different versions of the data", then for Instance Variable, there can be many (1..*)....? Not sure how a placeholder can have (?) multiple values....

Jenny Linnerud Dan Gillman could you help us understand this?

Hi,

I think there is a lot of confusion between the roles an Instance Variable can play for a particular Data Point and what the Data Point contains. Jenny Linnerud can correct me if I have this wrong, but this is my understanding. Lets consider 6 Instance Variables and their Value Domains (listed underneath):

IV1 - unitID

3 digit positive integer

IV2: gender

<m, male>

<f, female>

IV3: marital status

<m, married>

<s, single>

IV4: weekly wage

non-negative currency value (in Euros)

IV5: average weekly wage

non-negative currency value (in Euros)

IV6: GxMS cross product

combinations of gender and marital status

Consider the following microdata file for 2 units

ID    Gender   Marital Status   Weekly wage

001     m                 m             Euro 1500.00

002     f                   s               Euro 925.00

In this table, every cell is a Data Point, and the symbol written represents a Datum. For the Weekly Wage column, each Data Point has the following roles for some IVs:

IV1 - identifer

IV2 - attribute

IV3 - attribute

IV4 - measure

It is the measure role that supplies the Datum, the other IVs are descriptive in some way. The values are metadata for the cell.

For the Gender column, we use the following roles

IV1 - identifier

IV2 - measure

The other Instance Variables don't have a role.

The Marital Status column is similar.

For the UnitID column, the case is more interesting:

IV1 - identifier

IV1 - measure

Here, the unitID Instance Variable takes on 2 roles. We could also give attribute roles to the Gender and Marital Status IVs. That depends on what the application might need. There are no absolutes about these roles, except the measure role supplies the Datum as I understand it. There can be more than one measure contributing to the Datum in each cell.

For a Cube, the cells have the following roles attached:

IV6: identifier

IV5: measure

Note, we had to create a new variable to account for the categories in the Cube cells, because IV2 and IV3 only assign their categories to individual units, not the aggregates.

With more data, we might also identify one or more IVs with the attribute role for each cell. Each cell is identified by an element of the cross-product of the Gender and Marital Status value domains, producing 4 cells in this case

male, single

male, married

female, single

female, married

The Datum in each cell is the Average Weekly Wage, IV5. Here, the unitID IV does not take on a role, as it is not included in the aggregations.

A table representing the Cube might look like this:

                                                               Average Weekly Wage (in Euros)

Male                                                                            

     Single                                                                     1002.78

     Married                                                                  1498.03

Female                                                                        

      Single                                                                      951.22

      Married                                                                  1349.74

Note - these data are purely fictitious. (This proves data are not necessarily facts.)

Let me know if this is wrong or not clear.

Yours,

Dan

Hi,

Thank you very much Dan Gillman and InKyung Choi for clarifying it for me.

I see that I mixed up the type of a DataStructure component and the role of the Instance Variable.

I have to admit that my example was not that well formulated. My intention was that I have a group of 23 individuals and that I made a cube by aggregating different data of these individuals by gender and age. Gender and age would uniquely identify each aggregate. In Dan Gillman cube example he had an IV with the cross product. Would I have to create a similar IV or is there a way to express it as in my example?

Kind regards

Petros

Following Dan’s example, Dan Gillman can correct me if I have misunderstood,

IV2: gender

<m, male>

<f, female>

IV7: age category

<-20, less than 20 years>

<20+, 20 or more years>

IV8: average grade

Number between 1 and 5

IV9: gender x age cross product

Combinations of gender and age category

For a Cube, the cells have the following roles attached

IV9: identifier

IV8: measure

A table representing the Cube might look like this:

Male                     Average grade

                -20         3.1

                20+        3.2

Female

                -20         3.3         

                20+        3.0

Now back to the Data Point

Each cell in the Cube is a Data Point: (male, -20), (male, 20+), (female, -20), (female, 20+)

with Average grad as the datum in Data Point i.e. 3.1, 3.2, 3.3, 3.4 respectively

Yes, and rather than defining IV9, you could create 2 variables, one that maps age categories to cells and the other that maps gender categories. You have less control this way, though. The cross product idea is more sound and provides each combination correctly.

In reality, the idea of using the roles for cube data points doesn't really make sense except for the measure role. The cells in a cube are DEFINED and take their semantics from the elements of the cross product of all the dimensions that define the cube. The elements of each dimension are the categories from the value domains in the classification variables in the microdata. Looking at the definitions of the cells from the point of view of the dimensions rather than variables is cleaner. You don't have to go through gyrations to define the special purpose variables. Plus, dimensions can be reused as needed, and they already had to be created as enumerated value domains in the microdata.

The notion that the category combination identifies each cell is incomplete. The categories provide the semantics for each cell, and arguably this is the more important feature.

This is how we view the situation at BLS. We've built a model for cubes and time series that depends completely on dimensions. The only variables we care about are the measures. In addition, this model was incorporated into DDI and will be released as part of the DDI-4 Core product in January. The model has also been adopted by the OMG FIBO (Financial Industry Business Ontology) project.

I can share this model but it is in Visio, and I can't export it except for taking screen shots or cutting and pasting the diagrams into, say, a Word file.