Units in OECD.Stat Bo Sundgren

1
Units in OECD.StatBo Sundgren Lars Thygesen
2
Standardising decentralised databases

• Often based on paper publications
• Squeeze in as much as possible
• Mix apples and pears
• Generalised format The hypercube
• Obey certain rules
• Clean dimensions
• One (or more) parameter(s) (measure)
• Time
• Country

3
Example

• Residents in OECD countries 1990-2005 by Country,
  Time, Sex, and AgeGroup Count, SumOfIncome, and
  AverageOfIncome
• Dimension 1 Country (Member countries of the
  OECD)
• Dimension 2 Time (1990-2005)
• Dimension 3 Sex
• Dimension 4 AgeGroup
• Dimension 5 Parameters (Count, Sum(Income),
  Average(Income))

4
Organisation of OECD.Stat

• A number of relatively independent datasets,
  multi-dimensional tables
• common dimensions (e.g. country, time,
  frequency, age) provide links between the
  datasets
• compatible with Statistical Data and Metadata
  Exchange standards (SDMX)
• hierarchical thematic structure

5
Irregular statistical hypercubes

• Several classification variables concatenated in
  one dimension
• Different parameters for different dimension
  members
• or worse

6
As a result unit can be

• common to a dataset
• in a dimension
• differ for different members of one dimension in
  a dataset
• differs for different combinations of members of
  more than one dimension in a dataset
• not mentioned at all

7
Possible solutions

• The radical Restructure 500 datasets to become
  regular
• The pragmatic Keep datasets and repair

8
Proposal for a solution

• Introduce common Unit code
• Attached to any observation in any dataset
• Unit multiplier code
• Unit position code
• Dataset with only one unit
• Dataset with unit in a separate dimension
• Dataset where unit depends on only one dimension
• Dataset where unit differs for different
  combinations of members of more than one
  dimension
• Map unit texts to codes

9
Translation table
10
(No Transcript)
11
i - Unit of measurement Million USD
12
i - Unit of measurement Million USD
13
i - Unit of measurement Million USD
- Unit of measurement depends on the dimension
Country
14
i - Unit of measurement Million USD
i - Unit of measurement Million national currency
15
i - Unit of measurement Million USD
i - Unit of measurement Million national currency
- Unit of measurement depends on the dimension
Variables
16
i - Unit of measurement Million USD
i - Unit of measurement Million national currency
- Unit of measurement depends on the dimension
Variables
17
Units of measurement Click i for member of
dimension Subject to see unit
i - Unit of measurement Million USD
i - Unit of measurement Million national currency
- Unit of measurement depends on the dimension
Variables
18
Units of measurement Click i for member of
dimension Subject to see unit
i - Unit of measurement Million USD
i - Unit of measurement Million national currency
- Unit of measurement depends on the dimension
Variables
19
Units of measurement Please see explanation in
metadata i
Units of measurement Click i for member of
dimension Subject to see unit
i - Unit of measurement Million USD
i - Unit of measurement Million national currency
- Unit of measurement depends on the dimension
Variables
20
Problems

• Keep Unit clean as separate from measure
• Agree on unit codes (adapt to SDMX)
• Dataset owners think that even pragmatic
  solutions are too radical

21
Tentative time line
22
Questions

• Are these problems recognizable?
• Are similar solutions applied?

23

• There is nothing more practical than a good
  theory D. Hilbert