W w w. n e s s t a r. c o m Put your tables on the Web Jostein Ryssevik IASSIST/IFDO 2005 Edinburgh, Scotland Tables with drill down Hinged tables Dynamic.

w w w . n e s s t a r . c o m

Put your tables on the Web

Jostein Ryssevik

IASSIST/IFDO 2005

Edinburgh, Scotland

Tables with drill down

Hinged tables

Dynamic tables

Statistical tables

w w w . n e s s t a r . c o m

Topics

• Aggregated (tabular) data versus micro-data• Basic properties of aggregated data• Metadata standards for describing tabular data• The tabular data implementation in Nesstar• Demo

w w w . n e s s t a r . c o m

Micro-data versus aggregated data

Data about atomic units

Can’t be further disaggregated

Will normally be aggregated or handled by other types of statistical processes to generate analytical results.

Micro-data

Aggregated/tabular data

Data about aggregated units

Data are typically produced by an aggregation process from lower level data (I.e. crosstabulation)

…although some data might be global, only relevant to the specific aggregated level

Aggregation

Count

Sum

Average

Min/max

St.dev.

…or a more complex aggregation function

w w w . n e s s t a r . c o m

Relationships – from simple to complex

Aggregation

Micro-data Aggregated/tabular dataA table aggregated from a single source

Aggregation

A table where the nominator and

denominator of each single aggregated

value come from two different sources.

A table where each single aggregated data value is coming from a

different source.

w w w . n e s s t a r . c o m

Aggregated data - features

Aggregated data are typically multidimensional – an n-dimensional cube of data

Geography TimeSexAge

…and you would like to change the dimensionality of the table – know as dicing

Age SexGeography Time

w w w . n e s s t a r . c o m

Aggregated data - features

It is also nice to be able to slice the data cube – cut trough the data cube on one or more of the dimensions. In the example below we have sliced the cube on the time dimension, now only showing data for 2002

w w w . n e s s t a r . c o m

Aggregated data - features

The dimensions of a data cube will typically be hierarchical.

Drilling down a dimension

Rolling up a dimension

w w w . n e s s t a r . c o m

Aggregated data – an overview

• ….typically derived from one or more micro-data source though an aggregation process

• ….has a logical structure which is quite different from micro-data

• ….should not be understood as a “table”, rather as a dataset from which hundreds of different table displays can be derived

• Aggregated data (as most data) are best understood by separating the logical structure of the data from storage as well as presentation.

w w w . n e s s t a r . c o m

A three-layered model

Logical data model

(hypercube)

Presentation

Storage

DDI

Dimensions and measures described by the var-element. How

they are brought together in a

hypercube is described in the nCube-element..

A fairly generic mechanism to map the

logical model to the physical storage of the data is provided by the

location map

SDMX

Dimensions and measures and how

they are brought together in a

hypercube is described by the key-family

concept..

A physical transport/exchange

format for multidimensional data is offered (less generic

than in the DDI)

TabsML

A generic standards for describing the

presentation of survey tables (no logic).

Xcube

An XML standard for describing the logical structure of an OLAP

cube.

w w w . n e s s t a r . c o m

The twin roots of the DDI cube specification

The Vorburg compromise

• a great achievement, but• have been lacking some of the logic needed to drive functionality-rich tabulation system/OLAP systems.• a rigid relationship to the surveydata-oriented study concept.

A requirement to be able to describe

statistical products (tables) on paper or file

as they have been delivered form

statistical offices and other data producers.

Archived stuff.

A requirement to be able to describe live

statistical data held in tabulation systems,

databases, data warehouses etc.

w w w . n e s s t a r . c o m

The Nesstar implementation• Nesstar is a general purpose system for publishing and sharing

statistical information on the Web.

• A specific feature of Nesstar is the support for different logical data models.

Hierarchical/relational

Multidimensional

Rectangular

w w w . n e s s t a r . c o m

Live tables are great

fun.