This document has been prepared by the Singapore
Department of Statistics. It has been issued as
As the National Statistical
Authority of Singapore, the Department of Statistics
(DOS) is responsible for the management of the
national statistical information system. We
collect, compile and publish time series information
covering a wide range of statistics on the economic,
demographic and social characteristics of Singapore.
The vast amount of data (almost 30,000 series)
collected calls for an efficient data management
system. In 1996, DOS acquired FAME, a dedicated
software package for storing, processing and
analysing time series. In the same year, we
also launched TREND, a user-friendly Windows
based client for the retrieval and management
of time series.
CAPTURE AND PROCESSING
The data received by the
DOS comes mainly in the form of tapes, computer
spreadsheets and hardcopies. The data that is
received on tapes is loaded onto a mainframe
via batch jobs. This data is processed and then
loaded into FAME dB's. Data in the form of spreadsheets
are transferred and imported into FAME dB's
using an interface utility program. The remaining
data received in the form of hard copy is inputted
directly into the FAME dB's.
All of the time series
data captured by DOS are stored on a Master
dB. Depending on the security rating of the
series, the series is then assigned a security
code. This security code determines the sub-database
that this series is to be posted to after it
has been checked for errors. To make the most
of our server resources, this posting is done
after office hours via an automated production
For data that is to be
released to the public, the data is posted to
a Public dB. At the scheduled date and time
for release, an officer submits a job to generate
the necessary data files from the public database.
The data files are then transferred to an external
server for users to download.
The client that the public
uses to extract time series data from DOS is
TREND. TREND is based on Dbank, a windows-based,
tree-structured time series data management
client. Currently Dbank is also being used by
the Claremont Graduate School, National University
of Singapore and Yale University. The Dbank
web site may be found at: http://sunflower.singnet.com.sg/~dbank/.
Within TREND, data is
organised into subject groups. Each subject
group is organised into frequency groups. A
group is simply a collection of series. For
example, to locate the quarterly Gross Domestic
Product data, the user would look for it under
the ?National Accounts? subject group and the
"Quarterly" frequency group. This hierarchical
structure is modelled after the Microsoft Windows
File Manager, which makes data identification
and retrieval simple and fast.
To download data, the
users first identify and select series that
they want. The user then clicks on a button.
The TREND client will dial up, log on and download
the requested data. The system will then check
for changes in the meta data. If there have
been changes to the meta data, an updated version
of the meta data is automatically downloaded
by the TREND client. The connection is dropped
after this. By keeping a complete set of the
meta data at the client end, selection of data
by the users can be performed completely off-line.
This saves on connection costs for users. At
the same it also allows deployment of a relatively
small modem pool.
The TREND client also
has many user-friendly features to display,
report and graph the time series after it has
been downloaded. For example, it allows users
to export data to a number of statistical packages
and spreadsheet programs. It can convert data
to lower frequency during the export process.
TREND also provides an option to save data series
selected at any session in a "tagged file".
The tagged file can then be recalled for use
in future TREND sessions.
In conclusion this paper
has given a broad overview of how time series
data is managed in DOS. It has shown how time
series data is captured, processed and then
disseminated to the public. Given the extremely
rapid pace at which advances in IT are made,
this model is constantly being re-developed
and refined. There are for example plans to
put TREND onto the Internet. Also the FAME system
which is currently sitting on an AIX platform
is being migrated to an NT platform. Such changes
would bring about greater transparency, as it
would result in easier, cheaper and more widespread
dissemination of information.
1: MAJOR GROUPS IN TREND
Balance of Payments
Construction and Housing
Commerce and Services
Transport and Communications
Financial and Insurance
This is the TREND startup window.
The 19 sub-databases are shown in the "Group" box.
To select data series, choose
the desired subject and frequency group by double
clicking at it. A list of matrices in the
group will appear. Click on the required matrix
number to display the data series associated with
the matrix. Select a series to download by
highlighting it and right clicking on the mouse.
Series that have been selected will appear in the
"Tagged Series" box. Click on the ftp icon
to begin downloading.
Data downloaded from TREND can
be exported to a variety of formats for use in other
statistical/spreadsheet applications. To begin,
click on the export icon to open the "Export" window.
Choose the desired data format, enter the name of
the exported file and specify the first and last
periods (bounds) of the series to be exported. To
export, click on the Do it! button.
Here is a sample printout of TREND
data exported to an Excel spreadsheet.