Dissertation | Conclusions < Revised technique > Bibliography
Auditing catalogue quality by random sampling
8. A revised technique for the catalogue audit tool
The catalogue audit tool is successful as piloted for evaluating the accuracy of
bibliographic records. It should be considered just part of a toolkit for testing
catalogue quality, along with such tools as the collection-to-catalogue test for completeness, focused checks of the validity and appropriateness of assigned name and
subject headings, evaluations of serials cataloguing and duplicate detection.
- Before starting the audit it is important to profile the collection and its cataloguing.
The audit evaluates accuracy in the library's own terms, so variations resulting
from known local practices are not penalised. At a higher level, knowledge of the
sources of records and of the different cataloguing rules and software that have
been used is invaluable when interpreting the results.
- Choose the fields to be examined. Material description, genre, language and
location are often inapplicable; other fields may need to be introduced or moved
according to their perceived importance.
- Decide whether MARC errors and minor errors in capitalisation, punctuation and
diacritics are important or can be ignored.
- A random sample of catalogue records must be obtained, either from the
computerised library system or by systematic sampling of the shelves or a shelflist.
The number of records to audit depends on the acceptable margin of error in the
result: for a margin of 4.9%, 3.5% or 2.5%, choose a sample of 400, 800 or 1500
items, respectively. (If the error rate in the catalogue can be estimated, however
roughly, then it is often possible to justify smaller samples using the formula in
section 4.2.4.) The sample size is independent of the size of the collection.
- Full records should be printed for annotation with errors. If this is not possible, for
example when the sample is taken directly from the shelves, then errors can be
recorded concisely using the codes on the worksheet; alternatively, the worksheet
can be copied as many times as there are records in the sample and the relevant
boxes ticked.
- Check the cataloguing of each item in the sample using the worksheet (8.1),
comparing the catalogue record to the physical item. If an item cannot be found
and is not on loan, note its classmark and search for it on a second occasion; if it
is on loan then recall it.
- Detailed results can be displayed in the table given on the worksheet, or entered in
a spreadsheet or database for further manipulation. The principal figure to
calculate is the proportion of records with an error anywhere in Groups 1 and 2
(access points); the proportion of records with any error is also worth computing.
Cite these figures with the margin of error chosen earlier.
8.1 Worksheet
Control number: ______________________________________
Date of cataloguing: ______________________________________
If the catalogue record lists no copies held for this item, record code Z or tick this box [ ]
Please tick a box for each field in the catalogue record, or note the codes against each record
(for example T2 S1 M3). For author and subject headings, consider only whether they match
the form given in the authority file.
| acceptable |
incorrect or missing |
not applicable |
code |
Group 1 |
|
Personal authors |
|
|
|
A |
Corporate authors |
|
|
|
C |
Subject headings |
|
|
|
D |
Group 2 |
|
Title |
|
|
|
T |
Series |
|
|
|
S |
Group 3 |
|
Material description |
|
|
|
M |
Statement of responsibility |
|
|
|
R |
Edition |
|
|
|
E |
Imprint |
|
|
|
I |
Physical description |
|
|
|
P |
Language |
|
|
|
L |
Genre / category |
|
|
|
G |
Location / branch ID |
|
|
|
B |
Notes |
|
|
|
N |
|
1 |
2 |
3 |
|
If there are other errors in the record (for example absence of uniform title)
then list them below but do NOT count them in the statistics.
Dissertation | Conclusions < Revised technique > Bibliography
Owen Massey McKnight <owen.mcknight@worc.ox.ac.uk>