INDEXING MACHINE -READABLE DATA FILES

FOR A

SOCIAL SCIENCE DATA ARCHIVES

by

Jacqueline McGee

Rand Corporation

"It is still true that the best retrieval system is the expert human mind. "

This paper was presented at the lASSIST Annual Conference, May 19-22, 1983, in
Philadelphia, Pennsylvania.

INTRODUCTION

In the recent past much has been
written and discussed about the pro-
blems of cataloging and bibliographic
control of social science data. Many
of these problems may have been re-
solved with the implementation of the
Angle-American Cataloging Rules II,

Chapter 9 (AACRII) and the MARC format
for bibliographic control (2). How-
ever, there are a number of reasons
these solutions may not yet be univer-
sally implemented.

For instance, the AACRII and the
MARC format may be very familiar to
library staff, but all archives are
not staffed by librarians. Many
archives are suffering from a shortage
of staff and financial resources.
Federal agencies produce a major por-
tion of the data archived and used for
secondary analysis and these agencies
are also financially depressed.
Researchers and programmers who use
these Machine-Readable Data Files
(MRDF) are not as aware of the pro-
blems related to the acquisition or
storage of data and their interests
do not necessarily correspond to the
interests of the data archivist.

Technological changes occur so fre-
quently procedures may become obsolete
by the time implementation occurs. And
finally, so many new commercial firms

are installing social science numeric
data bases online and the interests of
these firms do not lie in the same
directions as those of the data archi-
vist. To assist the novice who may be

overwhelmed by some of these problems,
it is the hope of the author this

paper will provide some examples of
simple record keeping.

Rowe and Byrum previously described
a user-oriented system for the docu-
mentation and control of MRDF (3).

This system was comprised of four parts.

First, a standard catalog entry, second,
a data abstract or description form,
third, documentation codebooks and
lastly, the records of physical and
logical characteristics of the data
set. It is not the purpose of this

paper to offer an alternative system
for the documentation and control of

MRDF but to provide a practical example
of implementing such a system. This

example will provide an illustration
for the person who has just received



responsibility for the safekeeping of

a collection of MRDF or to establish

an archive and isn't sure where to

begin. The first item in the system
by Rowe and Byrum, the standard cata-

log entry, was described before the

Anglo-American Catalog Rules II,

Chapter 9 were implemented. Data
librarians located in a traditional
library are already familiar with the

rules for cataloging, but may not be

familiar with the AACRII, Chapter 9.

The Anglo-American Cataloging Rules

II, Chapter 9 describes the standard
rules for cataloging MRDF. It is not
within the scope of this paper to

argue the pros and cons of the accept-
ability of the AACRII. There can be

no doubt that a uniform standard de-

fining MRDF is necessary in order to

alleviate present confusing practices
and the proliferation of titles for

one data file. Certainly implementa-
tion of the AACRII and the agreement
of the MARC format were giant strides
in the cataloging and bibliographic
control of MRDF.

STANDARD CATALOG ENTRY

Rowe and Byrum state "Standard
catalog entries, constitute the pri-
mary records by which computer-
readable data files should.be con-
trolled and accessed." It is with
this one area of their discussion that
I disagree slightly. The standard
catalog entry requires extensive
staff time and financial resources
and need only be considered as neces-
sary under certain conditions; if the
required resources are available and
may be allocated to such an endeavor;
if the data archive or data bank is

situated in a library or a library is

available and willing to participate;
if the data holdings are original
data from the institution responsible
for the establishment of the archive.

It is hoped non-originating
archived data will be catalogued by
the originating institution. However,
the federal government is responsible

for a major portion of the data files

held in many archives and current fis-

cal restraints on most federal agencies

probably will not permit such a project
in the near future. There is, however,
an ongoing cataloging project at

Michigan's Inter-University Consortium
for Political and Social Research
(ICPSB) which may resolve the problem
of cataloging federal data (4). ICPSR

is certainly one of the largest, if

not the largest, of the data archives
in the United States. When this pro-

ject to catalog their holdings is

complete, it may be possible to consi-
der a union catalog.

DATA ABSTRACT OR DATA DESCRIPTION FORM

It is the Data Abstract or Data
Description Form described by Rowe
and Byrum which should be given prior-

ity in the development of an archive
record system. The data abstract or

data description form in a standard
format is an absolute necessity and

should be the core of the documentation
for the archiving of MRDF.

Aldrich has proposed a similar
Abstract Form for the documentation
of federal MRDF (5). It was from a

description by Aldrich the following
example was derived. Changes made in

the form were for the benefit of the

user and do not reflect a disagreement
with her proposed standards. Since
this form includes an abstract summar-
izing the data set or file being
archived, this document shall be

referred to here as a Data Base Pro-

file. With some slight variations,
the items contained in the form gener-
ally should contain the items shown
on the next page.

If it is not possible or feasible
for the archive or library to catalog
the holdings of the archive according

to AACRII at least the information
supplied in the Abstract or Data Base

Profile form will conform to the stan-
dards for describing MRDF. If at

some future time cataloging is possi-

ble, the information for the catalog



Abstract Form for the Docwnentation of Federal MEDF

FILE #: an identifying number for the individual archive

FILE NAME: a title

FILE SOURCE: producer, distributor, processor

PRINCIPLE INVESTIGATOR: primary researcher

TYPE OF FILE: survey data, microdata, administrative records,
process records, geographic records, software

UNIVERSE: total universe the records describe

SAMPLE SIZE: number of observations or records

SAMPLE UNIT: household, person or unit being measured

RESTRICTIONS: none, or any restrictions placed on the distribution

ABSTRACT: a summary description of the data set or file.

Each abstract held in the archive should contain
the same information. Care should be taken not to
omit any portion of the required information and

the information should appear as closely as possible
in the same paragraph. This assures an easier
search for the individual . looking for specific
information as to size of the sample, purpose of

the study, key variables, etc.

REFERENCES: a descriptive listing of the hard-copy documentation
available for use with the data file, i.e., code-
books, survey instruments, dictionaries, etc.

RELATED PRINTED REPORTS: known reports where the data is described or where
the data file has been used

TAPE SPECIFICATIONS: the physical characteristics of the data file, the

tape numbers, the logical record length, the block-
size, data set names and density



entry will be readily available.
Copies of the Data Base Profile

may be stored in computer format as

well as in hard-copy. If the data is

stored in computer format it would be

possible to devise a simple online
search capability. If the data libra-
rian wishes to be bibliographically
correct, study the AACRII and include
in the Profile the pertinent informa-

tion from the AACRII as well as the
information required or deemed neces-
sary for the institution housing the

data library (7). Many of the ele-
ments of the Data Base Profile may be

utilized to produce a catalog. For
instance, each abstract when extracted
from the Profile provides summaries
of the archive holdings. The Rand

Data Facility catalog uses the

abstracts in such a way.

Each abstract then written, there-
fore, should include the following

(6):

• Data base identification number

• Source

• Name

• Date the information was collected

• Subject

• Geographic level of the data

(lowest)

• Population or sampling unit

• Number of observations or number
of logical records

• Key variables

Indicies may then be derived from the
information given on a Data Base Pro-

f i Te

.

SOURCE AND NAME INDEX

The Source and Name Index is de-

rived from the File Name and the File
Source as given in the Data Base Pro-
file. Often these items are sorted
as separate indices; an author index
and a title index.

An index should lead a user to the
information he is seeking with as

little effort as possible, and so we

have combined these indices.

On the Data Base Profile and in

data file records we use the most
correct name for a data file. The

correct name may be drived by using

the AACRII rules. Since we also wish
our individual indices to assist the

user in his search we also include in

our local archive index those aliases

or acronyms when they are commonly
used, if the index is to prove useful.

In order not to have a great many
"see ." included in the index
where aliases or acronyms or common
usage names are listed, the correct
identifying number of a particular
data file is used as the pointer.

KEYWORD INDEX

This index is certainly one of the

most difficult to construct. A the-

saurus would be helpful; however, the

keyword index discussed here was deve-
loped from individual data files. As

mentioned earlier, one of the mandi-

tory sections of the Data Base Profile
is a list of key variables. At the

time of archiving a new data file, the

abstract is written and the key vari-

ables extracted. These variables or

subject categories are added to the end

of the keyword index. A copy of the

keyword index is kept online. New key-

words are added at the end of the old

index and a short SAS program sorts the

keyword index by words or by identify-

ing file numbers whenever necessary.

GEOGRAPHIC LEVELS AND MAJOR SUBJECT
VARIABLES IN THE KEYWORD INDEX

Using Census Bureau designations,
the lowest geographic level of the

data is assigned as a second element

of the keyword index. By using the

lowest geographic level it is possible

when searching for data to weed out

those data files not useful to the

researcher.
A third element for the keyword

index is a prescribed list of major
subject variables. For each data file

the keyword index will include at least



one, but not more than three major

subject categories. It is then possi-

ble to produce tables listing data

files by geographic areas and major
subject categories. Librarians may

want to use the Library of Congress
Subject Headings (LCSH) List.

The Data Base Profiles may be

produced as printed copies to be

given to researchers interested in

using a particular data file and can

be used as documentation for bibli-
ographic citations in research papers
for the creating of a catalog of

holdings.
The Data Base Profiles stored in

a partitioned data set at Rand are

soon to be converted to a total on-

line system using the IBM INFO/SYS
CSD data retrieval system.

The OZ INFO/SYS has the capability
of handling multiple data bases and

has been utilized at Rand recently to
create an information system for a

library of software models as used in

one department (8).

With the data stored in OZ it will

be possible to search the system for
a particular data base by title, by

keword, or keyword combinations. Re-
questing data bases by keywords pro-

duces a "hit list" of all data bases
that contain the requested keywords.
The hit list can then be accessed
in either of two forms: one is an

abreviated form that lists only a

few specifics about the data bases
on the list (including the data base
number and title); the other is a

full description of the data base.

It is possible to get print copies
of an OZ screen, of the individual
data base descriptions and of the
hit list of keywords.

References

(1) Michael Phillipson, "The Classification, Storage and Retrieval of Survey
Data, Reader in Machine-Readable Social Data,: H. White, ed. , Information
Handling Services, Englewood, Colorado.

(2) See Sue Dodd, "Cataloging Machine-Readable Data Files: An Interpretive
Manual," American Library Association, December, 1982.

(3) John D. Byrum, Jr. and Judith Rowe, "An Integrated, User-Oriented System
for the Documentation and Control of Machine-Readable Data Files," Library
Resources & Technical Services 16(3), Summer, 1972.

(4) Carolyn Geda, "Marc Formal Applied to Machine-Readable Data Files: A Pilot
Project of ICPSR," Paper prepared for delivery at the annual meeting of the

Society of American Archivists, September 1-4, 1981, Berkeley, California.

(5) Barbara Aldrich, "Proposed Standards for Bibliographic Entry and Abstracts
for Federal Machine-Readable Data Files," Redraft 2, October, 1978.

(6) Don Trees, "The Rand Computation Center: Rand's Data Facility: A Guide to

Resources and Services, " BCC-1555/18, Santa Monica, April, 1981.

(7) Anglo American Cataloging Rules II, Chapter 9, Machine-Readable Data Files,

P201-216.

(8) As described by William Fowler, The Rand Corporation, Santa Monica,
California, 1983.

10