Message-Id: <3.0.1.32.19981117101651.00987890@mail.trd.sintef.no>
Date: Tue, 17 Nov 1998 10:16:51 +0100
To: MedSoc@csf.colorado.edu
From: Aksel Hn Tjora <Aksel.H.Tjora@unimed.sintef.no>
Subject: Re: Software for Qualitative Data
Mime-Version: 1.0
Content-Type: text/enriched; charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable
The use of software for qualitative data analysis is very useful.
Especially the packages are effective in the process of indexing on line
data, as for example field interview transcripts. Jesus Ramirez Valles
says that he is  afraid that the software may take over the analysis.
Software, like NUD-IST does not require any alternative form of data
analysis than the manual method, I think, but Valles is right in a way
because one is tempted to use the default procedure as presented for
example in the software manuals. NUD-IST asks you to make a hierarchical
tree of indexes, which may not fit the way your indexes are related.
However, you may design that index tree as a one level flat tree, and
thereby work around this software determinism. Another problem that I
experienced with NUD-IST was that changing the indexes during the
research project would cause quite a lot of work. This means that you
have to work rather hard with index construction in the beginning of your
data analysis to avoid fundamental changes later. This problem made me
put NUD-IST aside in my thesis work on medical emergency centres, because
I had to change my indexes drastically in a rather late stage, and I did
not have the energy necessary to work through the indexing again. The
result was that I used NUD-IST as a sorting tool for my data, rather than
an analysis tool. I think one should be rather concerned about the
dynamics in the indexing when choosing a software, as I find it necessary
to be able to change indexing, heavily if necessary, during all stages in
the research work. Check out the page at Sage Publications for various
software and downloadable demo versions, at http://www.scolari.co.uk/


The following subsection from my PhD thesis (in sociology) describes in
short terms my experience with NUD-IST:


--------------------------------------

<bold>Using NUD=B7IST in Qualitative Data Analysis

</bold>After the three first periods of interviews and two observation
periods, I used the computerised tool NUD=D7IST (Non-numerical Unstructured
Data Indexing Searching) for sorting data. The program is organised into
a document part and an index part, and applies a cut-copy-paste process
on the raw data to construct a very flexible way of relating text pieces
or other data formats to an index system. The index sys-tem is
hierarchically ordered. Besides, there are well developed search engines
for searching the raw data, for indexing, or searching the in-dex
systems, for finding clusterings, testing hypothesis, making quali-tative
matrices, etc. (QSR, 1994).=20


After being introduced to NUD=D7IST by Amanda Bow and Karen Wale in March
1995, shortly before the first interviews of nurses in LV cen-trals, I
decided to use NUD=D7IST in my data analysis. Although I had planned to use
NUD=D7IST for the complete analysis, I found it difficult because of the
hierarchical structure of categories (indexes), that NUD=D7IST is based on.
I have problems with accepting the idea of the designers Richards and
Richards (1995), that it is natural to think in terms of hierarchical
ordered categories. They think of the categories as sorted by a
general-to-specific process (Richards and Richards, 1995:82). A source of
the problem may be that the categories are thought of as characteristics
(like age, gender and religion), a way of thinking that stems from survey
approaches.


My analysis focuses mostly on the relations between the people,
institu-tions, professions and technologies that I study. Following the
catego-rising strategy of NUD=D7IST in the complete analysis, I feel that I
would have to separate these units, and their work and practice, and the
con-textual factors. I am afraid that it would draw me too far away from
the relational nature of my data. As pointed out by Kelle and Laurie,
implementing a consistent and stable coding scheme at a too early stage
of data analysis could have a "hazardous influence on hypothesis
generation" (Kelle and Laurie, 1995:28). I believe that the closeness to
the data is needed in the exploring strategy that is followed. In a
re-view of NUD=D7IST, Weitzman and Miles pointed out some of the same
drawbacks with the program as myself; "distance from the data,
con-text-less search hits, [..] and hierarchical-only coding" (Weitzman
and Miles, 1995:256).


Nevertheless, NUD=D7IST provided a very useful way to sort the data, so
that the sorted lot could be retrieved at any moment. As the amount of
data grew, it was easy to include into the NUD=D7IST document system, and
browse the raw data, addressing fragments to different categories
(indexes). The hierarchical index tree was modified continuously, as new
analysis of new data made me think of new categories, or as analysis made
it natural to delete or merge old indexes. After three in-terview periods
and two observation periods, I had a large amount of data, that was
systematically ordered. However, that the ordering of the indexes is
hierarchical was more confusing than useful for this preliminary data
analysis of sorting the observed phenomena. Hence, without using the
index tree, I still used the indexes. I left the NUD=D7IST approach after
having established the preliminary categories, and pursued a manual,
intuitive sorting and categorising approach of the data from the
observations and interviews that came later.=20


The use of NUD=D7IST is not only interesting in a methodological
discus-sion, but also in a discussion of technology and work practices,
hence in the focus of the thesis in general. When I started to use
NUD=D7IST, the program was of course already fully  designed. Nevertheless,
I was able to reconstruct the designers' idea of a hierarchical ordered
index tree, to a bunch of indexes that for me at that time were
unre-lated. I put all indexes on one level of the tree, so no index would
be subordinate to the other. My use of NUD=D7IST was therefore not really a
use of an analysis software, but a use of the software reconstructed as a
very useful tool for selecting and ordering raw text by categories.


In the same way as many of the LV nurses more or less reconstructed the
LV centre, that was supposed to be a service telephone into a
switchboard, I appropriated NUD=D7IST to be a sorting tool, an advanced
file manager. Although the artefact was already designed, it did not stop
me from redesigning the use of the system to suit my own pur-poses, or
avoid using the parts of the software that were conflicting with my
research strategy and aims.=20


As pointed out by Coffey and Atkinson (1996), it would be wrong to allow
the software to drive the research strategy. After having used the
functions of NUD=D7IST that (I think) did not alter my research strat-egy,
I felt it was right to leave the software for more manual methods.

---------------------------------


Hope some of my experience might be helpful.


Aksel Hn Tjora

<color><param>8080,8080,0000</param>_____________________________________

Aksel Hn Tjora (akselt@unimed.sintef.no)

Researcher, Associate Professor, PhD

Norwegian Institute for Hospital Research (NIS)

Address: Sintef Unimed, 7034 Trondheim, Norway.=20

Tel: (+47) 7359 7889   Fax: (+47) 7359 6361 Tuesdays

Tel: (+47) 7359 0267   Fax: (+47) 7359 1564 Other week days

Private tel: (+47) 7352 8552</color>