48 - iassist quarterly

The Development of a

Canadian Union List

of Machine Readable
Data Files (CULDAT)

This article is an abridged version of the final

Report, "Pilot Project for the Development of a

Canadian Union List of Machine Readable Data

Files (CULDAT)," prepared by Edward H.

Hanis, Social Science Computing Laboratory,

University of Western Ontario for the Machine

Readable Archives, Public Archives of Canada.

A survey of Canadian social scientists,

undertaken in 1982, indicated that a need

existed for an inventory or union list of data

files available for secondary analysis. In the

mid-seventies, the Data Clearing House for the

Social Sciences (DCHSS) had spent considerable

time and effort in the development of an

automated inventory. The loss of DCHSS, due

to lack of funding, unfortunately also involved

the physical loss of the magnetic tape which

held the descriptions of these files. The results

of the 1982 survey indicated strongly that the

research community continued to feel thai a

union list of data files would be a valuable

resource. In response to this need, the Machine
Readable Archives Division (MRA) [of Public

Archives Canada] established a contract with the

Social Science Computing Laborator>' of the

University of Western Ontario to develop an

online inventory describing computer files held

by Canadian data archives and libraries. The

overall purpose was to develop organizational,

technical cind informational foundations for

maintaining and disseminating a computerized

inventory. Specific objectives involved: the

establishment of a standard for describing

MRDF for entry into the data base; the design

and implementation of the pilot data base

containing a partial inventory; and the definition

of the organizational roles and mechanism to

effect routine and cost-effective flow of

descriptive information from data archives and

other organizations to the imion list beyond the

conclusion of the piloL

The pilot project was carried out over a

fourteen-month period. In January of 1985, a

committee of data archivists and data librarians

established a list of elements which were to be

used to describe the holdings of the institutions.

These elements were taken from those defined

in the MARC format for data files. A data

dictionary was developed to aid participants in

the entry of descriptive information. The Social

Science Computing Laboratory was involved in

six major activities: the aeation of the pilot

data base; the set-up of online access with

Basis on the lab's VAXll/785; the set-up of

DATAPAC and standard dial-up

communications; conducting an evaluation of the

online system; a survey of potential

contributors; and production of a hard copy

reference doctimenL

Contributors to the data base were from the

university-based archives and libraries and

included: Data Library, University of British

Columbia; the Institute for Social Research,

York University; Data Resources Library,

University of Western Ontario; Institute for

Social and Economic Research, University of

Manitoba. The MRA also contributed

descriptive entries. In all, 753 records were

entered into CULDAT. Evaluation of the data

base was extended to more participants than

those listed above and included both frequent

users of online systems as well as infrequent

users. Although a number of suggestions have

Summer 1986



iassist quarterly - 49

been made as tohow to improve the online

inventory, the general consensus was that the

data base was very useful and shotild be

continued.

It is not siuprising that the most crucial

component of the data base was the description

of the data file. A number of difficulties were

experienced with the lack of consistent

terminology used and the detail of the

description itself. The problems encountered are

simunarized in the following paragraphs. The
resolution of these difficulties have formed the

basis of the CULDAT work plan for 1986-87.

The choice of data elements to be included in

CULDAT was based on the fields in the

MARC format for data files. A limited number
of elements was chosen as it was felt by the

conomittee that the intention of the data base

was to include only sufficient information to

identify a imique data file, to aid researchers in

selecting files of interest, and to locate archived

copies of the file. The resulting CULDAT Data

Element Dictionary contained the field names
and a brief description. During the pilot

project, it was noted that in some cases the data

dictionary did not provide sufficient guidance to

the archivist or librarian to allow him to

adequately describe data files, and presumed a

knowledge of the MARC format and

Anglo-American Cataloguing Rules II. This

CTeated some difficulty in mapping out the

information received for input into CULDAT.
The consequences of a weak data element

dictionary are inconsistent presentation of the

information which can make the descriptions

difficult for the end user to interpret Weak
data descriptions yield inefficient indexes, which,

in turn, require that the user anticipate all

possible variations of a term in order to find all

relevant records in the data base. Specific

problems were fotmd in the following data

elements.

1. Investigators : The differentiation

between principal investigator and other

investigators caused some difficulties for

both cataloguers and the users. The
determination of principal investigator for

a data file is difficult, if not impossible,

at times. The separation of these fields

requires searching two fields rather than

one for the user wishing to browse the

index. The distinction between

investigator (personal) and investigator

(corporate) was considered essential. The
lack of authority control in the corporate

investigator field was a problem which

could be overcome through the use of

Canadiana to control the use and spelling

of names.

Producer: Generator. Distributor : A
tendency to repeat the same data in these

fields was found. This may have been

due to the inadequacy of the data

dictionary. Abbreviations and acronyms

were used. The adoption of an authority

file for corporate names would apply to

these fields as well.

File Size: Number of Cases : Some
difficulty was experienced in the data

provided in this field. Again, this was

due to lack of guidance in the data

dictionary.

Access Restrictions : As all institutions

have their own access regulations, it was

felt that this field shotild only be

completed when the distributing

organization has contributed the record.

Abstracts : Information contained in this

field was found at times to repeat

information found in other fields. The
vocabulary used varied widely which

made control of the field extremely

difficult The types of variables used in a

data file is vital information for the

prospective user. In order to provide

improved access to this field, it would be

Summer 1986



50 iassist quarterly

preferable to separate the abstract from

the variable list Variables could then be

left unindexed. Such a change would

significantly reduce the indexing overhead

and improve the quality of the printed

keyword index by using variable names

instead of individual words. The online

system could continue to index variables

as individual words as well as expressions.

Geographic Coverage : The pattern

adopted by the pilot was as follows: site,

city, region, territory, province, state,

country (quahfier) continent The pattern

worked well in most cases and ensured

that the user interested in data about a

particular province could retrieve

information on a file which covered only

a city in that province. The only records

which do not conform to this pattern are

physical data where orbital coordinates

are submitted.

Chronological Coverage : The format of

the dates recorded in this field was

inconsistent, rendering the retrieval of

data ineffective. The data dictionary

should prescribe one acceptable format to

which all dates would be converted. A
standard format will provide the

possibility of performing systematic

retrieval on time periods by scanning the

text, even though every unit of time

within a range is not actually recorded in

the field.

The difficulties which have been encountered

will provide valuable information to allow us to

improve the quality and guidance required for

the data dictionary. The second version should

improve the consistency of the descriptive

entries. The conuibutions made by the data

archives and libraries were extremely useful in

building the pilot data base and allowing us to

identify specific areas for improvement in the

data dictionary.

User Evaluation and Potential Contributors

The original project design called for online

testing and evaluation of the pilot CULDAT
data base by project participants and

constructing a list and contacting potential

contributing organizations in order to learn

about their holdings and interest in submitting

entries into CULDAT in the future. Three

important additions were made to enhance the

project The establishment of a DATAPAC
Service reduced usage costs and significantly

improved convenience to remote users. In

addition, the survey of contributors was

expanded to include questions on evaluation as

they were potential users as well. The third

activity was to include three local University of

Western Ontario groups (students in the School

of Library and Information Science, the

University's reference librarians, and social

science researchers who use the Lab's support

services). These additions increased the use of

CULDAT during the pilot phase.

The evaluation of the data base was very

favourable and many respondents expected to

benefit from the availability of CULDAT in the

future. Considerable information from

prospective contributors and users was acquired.

This information and experience provide a

sound foundation for the design and planning of

the next stages in the development of

CULDAT. The major activities planned for

1986/87 will include: 1) the revision and

expansion of the CULDAT Data Dictionary in

order to provide more guidance on the

description of holdings for entry into CULDAT;
2) continued support to university based data

archives and libraries to ensure their holdings

are included in the inventory; and 3) the

redesign of the formatted hardcopy version to

make it available as a reference document at

less costn

Summer 1986