Subject: codes and indexes |
From: Herve Varenne <hhv1@columbia.edu> |
Date: Tue, 09 Sep 2003 18:24:41 -0400 |
To: Frederick Erickson <ferickson@GSEIS.UCLA.EDU>, CAELIST@listserv.vt.edu |
Subject: Re: Analytic Induction // Quantitas-Qualitas |
From: Jon Wagner <jcwagner@UCDAVIS.EDU> |
Date: Tue, 9 Sep 2003 14:01:53 -0700 |
To: CAELIST@LISTSERV.VT.EDU |
RE: Analytical Induction
Howard Becker's book Tricks of the Trade has a nice short section on
analytic induction -- and compares this to several other forms of
inductive social analysis as well. You can find the section by using
the index, but the whole book is chock full of examples and guidance
about doing just this kind of work.
There's also a very good chapter in the beginning of of Boys in White (by Becker, Blanch Geer, Anselm Strauss and Everett Hughes) called" The Design of the Study" in which the authors state up front that," In some sense, our study had no research design." However, rather than presenting that as an anti-method throw-away line, they then provide a detailed account of how a design for the study did emerge from their initial interests and their efforts to be systematic and conscientious in how they collected and analyzed data about the people they were studying (medical school students).
RE: QUANTITAS/QUALITAS
As a nifty example of Fred Erickson's point about qualitas and
quantitas, the Boys in White authors also systematically compared (a)
all instances of having "overheard" a statement in natural
conversation with (b) all instances in which the statement was
volunteered in response to a researcher's question. They treated a
high ratio of a:b as a sign of confidence that the statement was
important/salient to the people they were studying, and a low ratio
of a:b as an indication that it might be an artifact of the
researcher's presence. In essence they did the math . . . but they
did it on statements that the people they were studying were making
to see if the researchers were getting even that right before moving
on to interpret what these statements -- when taken together -- might
mean about medical school.
My sense is that by providing more detailed, explicit descriptions of
analysis processes/rubrics such as these -- and why they are
necessary -- a very strong case could be made for "rigor" in
educational research that would neither disadvantage qualitative
studies nor advantage quantitative ones. We should count what it
makes sense to count, when it makes sense to do so. We should also do
our best to understand how the categories and "kinds" of things we
refer to (some of which we count) correspond or don't correspond to
what the people we are studying seem to care about -- not just in
research interactions, survey (test) responses, or narrowly defined
institutional settings, but in their lives. To my way of thinking,
that's what "rigorously empirical" OUGHT to mean.
Jon
I was to first express my delight that this conversation about qualitative software is taking place. This is a question that arises frequently in my interdisiciplinary work setting. In particular, I was elated to read Fred Erikson's suggestion not to use qualitative software to analyze ethnographic data. Not only does much of that software lead to premature categorization and interrupt that iterative process Fred describes so well, it also discourages other modes of interpretive inquiry such as textual or narrative analysis. Though it is very time-consuming, I have found that nothing replaces the insights gained by systematically and critically reading and rereading multiple sources of data in relation to each other. Even in larger team projects, much can be gained through a comparative reading of each other's interviews, observational notes, etc.
However, I have had a hard time making this case to some of my colleagues, in large part because they view qualitative data as "containing information," rather than as being comprised of "texts."
I was wondering whether Fred or others could suggest texts or articles that discuss the process of "analytic induction."
Thanks!!
--Janise Hurtig
Subject: Re: recommendations regarding qualitative software |
From: Mica Pollock <micapollock@MINDSPRING.COM> |
Date: Mon, 8 Sep 2003 18:47:23 -0400 |
To: CAELIST@LISTSERV.VT.EDU |
Just a clarification to this nice discussion: why not teach the iterative analytic/annotating strategies below AS the first (main!) processes OF good "coding," rather than eliminating the word from methods courses? They'll be told to "code" in all their other classes, so perhaps it's best to make clear with such analytic exercises that when anthropologists "code" we strive for systematic understanding while resisting too-easy, self-imposed, or too-early analyses. . . Just a thought on language, given the current climate re. ethnography, etc. Mica Pollock
Subject: Re: a P.S. on "qualitas" |
From: Frederick Erickson <ferickson@GSEIS.UCLA.EDU> |
Date: Mon, 8 Sep 2003 18:08:33 -0700 |
To: CAELIST@LISTSERV.VT.EDU |
In my previous message I had meant to say something in a bit more explicit way and then rushed into "Send"--so easy to do with e-mail. So I've elaborated a bit in this postscript. It responds in part to Mica's wise observation that in these times of hostile U.S. federal climate for "ethnography" we need to be able to show how what we do can indeed be systematic. (Maybe we should even call what we do "coding", as protective coloration.) So here's a few more lines on qualitative foundations for "careful coding" and quantification. Here it is. It wasn't until two years ago (I'm embarrassed to say) that I realized that the Latin contrast terms QUALITAS/QUANTITAS made a distinction that often is overlooked, or misunderstood; one that is rhetorically useful for qualitative researchers. "Qualitas" refers to the KINDS of things there are in the world (or are in a certain domain of scientific inquiry), and thus in interpretively oriented ethnography it refers to those KINDS of things (and KINDS of KINDS) that are relevant to social actors for the conduct of their everyday lives--more formally, to the folk ontology (or ontologies) extant among a particular set of persons who interact routinely. So the primary/generic research question in qualitative research is "What are the kinds of things that are important for the conduct of social action in this local community of social practice?" "Quantitas" refers to the AMOUNTS of things there are in the world--thus the primary question in a quantitative study is "how many of something are there in this place in the world?" A point that quantitative researchers either miss entirely, or under-recognize, is that when one asks as a guiding question "how many?" what one ends up counting are INSTANCES OF KINDS OF THINGS--i.e. there is a foundation of qualitative judgments about what's there in the world that's of research interest (and countable) which is prior to and thus underlies any and every quantitative analysis. When you are counting instances of the right KINDS of things, hermeneutically--not just in terms of some substantive theory or other, but in terms of the points of view of the social actors whose actions/opinions you are counting, then your quantitative analysis can produce valid conclusions. But when you haven't got the qualitative foundation/taxonomic framework right--e.g. you are counting what from members' points of view are apples and oranges but as a researcher you are calling them all apples--you can never reach conclusions that have validity--at least not "interpretive validity" concerning social action, for meaningfulness to the actor is always entailed in social action as a notion. And so, the problem with naive use of the "coding" software is that it asks the researcher to make snap judgments about QUALITAS and then immediately to start tabulating QUANTITAS. That's why "premature typification" is so dangerous and misleading. There's nothing wrong with counting things (or, to my mind, with typification itself)--but you need to take great pains over time in order to discover the right things to count. Counting (even fancy counting, as in inferential statistics) is the easy part. Discovering the right kinds--the right entities--to count is the hard part. We need to be clear about this, among ourselves and with our quantitatively oriented colleagues. I love it when ethnographers count things--if and when they are very careful about what to count and can show me why they are counting what they do count. Frederick Erickson
Subject: Re: Recomendations regarding... |
From: James Mullooly <jmullooly@CSUFRESNO.EDU> |
Date: Mon, 8 Sep 2003 09:57:33 -0700 |
To: CAELIST@LISTSERV.VT.EDU |
Kevin, Although students may fall prey to "premature categorization", such software is great for indexing all of your data. In this way, you can actually find something you wrote about an observation (and forgot about) very quickly. I used Atlas.ti < http://www.atlasti.de/ > for my dissertation and it worked very well. It is a very powerful program that can do most of what most such programs do. But I would agree with Fred that Nudist < http://www.qsr.com.au/ > is a good way to go. Nudist can do most of the same things as the more powerful packages and it is better known. As more people are familiar with it, working with others is more likely and finding training sessions for it is easier. Although the 'learning curve' with these things can be steep, it was well worth it in my estimation. -Jim PS See also, Barry, Christine A. 1998: 'Choosing qualitative data analysis software: Atlas/ti and Nudist compared'. Sociological Research Online, vol. 3, no. 3. http://www.socresonline.org.uk/socresonline/3/3/4.html. ============================= James Mullooly, PhD Assistant Professor Department of Anthropology California State University, Fresno http://www.mullooly.homestead.com From: CAELIST Discussion List [mailto:CAELIST@listserv.vt.edu] On Behalf Of ken jacobson Sent: Sunday, September 07, 2003 5:15 PM To: CAELIST@LISTSERV.VT.EDU Subject: Re: Recomendations regarding... Hi, the following system worked with several thousand pages of field notes. I went through the first couple of notebooks and used colored tabs to try to find categories that seemed to fit the notation. The tabs got unwieldy so quickly, that I took my nascent categories and transferred them to big accounting pads. I gave each book a number and as I went through the notes numbered the pages. I then listed under a category (the numbers of which kept grown) on the accounting pad, the book, page and a brief synopses of the interesting snippet. I suspect there are computer spread sheets that would work if you're into typing. When it came time to actually use the data, I found the system worked better that I hoped. So the big deal for me is to find categories that seem to work (their names too can be changed) and then to look for emergent patterns both within and between categories. Good Luck!!! Cheers...Ken Jon Wagner wrote:
Let me build a bit on the comments by Joe and Fred -- by recommending (as Joe did) the Weitzman and Miles book as an excellent guide to thinking about different ways that you can use computer software and by recommending (as Fred did) that you not jump into coding data too quickly -- while at the same time recommending two software programs that you should consider, neither of which may be very well known. The starting point for me in all this is that I used to teach a section on "coding" in my qualitative research course, but I've dropped it for the very reasons Fred mentions: students were just too, too eager to take a few key terms and start applying them, to every phrase or paragraph or section in an interview transcript, or field note, that might fit. Sometime after that, I'd start getting questions about what kinds of theory they could use to analyze their coded data, questions that I had to answer as follows: "If your data is already coded, your theory is already set because, at the end of it all, a code is nothing more or nothing less than a set of terms that you use to articulate your data with your theory." Others may disagree, but people who share this notion of coding worry constantly about getting those terms fixed too soon. You can't begin a research projects without some idea of what you want to study, but you lose a tremendous amount of potentially valuable data every time you narrow your view of the world/phenomena/etc. to a particular concept, hypothesis, question, etc. That's not an argument for avoiding this kind of focusing, but it is an argument for delaying it long enough so that HOW you narrow your attention can be informed by observations and evidence about what you are studying. So what I teach now, instead of "coding," is "annotation," a process of making notes on and about observations, field notes, interview transcripts, video recordings, etc. This is an iterative process, because you the make notes about your notes as you go over the material again, and again, and notes on those notes, etc., all with the purpose of trying to understand what is going on in/with/at the place/people/situation you're interested in. And it's with this idea of "annotation" in mind -- which, after enough time, attention and analysis, can lead to an appropriate exercise of the kind "coding" that too many people are too quick to jump into -- that I'll mention two programs that have served me, and several of my students, very, very well. FileMaker Pro: If your data are in the form of computer text files -- interview transcripts, field notes, email postings, etc. -- you might take a look at a general purpose data base program that does a great job of handling text. FileMaker Pro is an excellent candidate, for this, but some people have also had good luck with Microsoft Access (including the Centers for Disease Control, where staff have developed some customized templates that you can download for free). What FileMaker allows you to do is to identify a chunk of text of any length -- a word, phrase, paragraph, page , whatever, up to about 64K (20+pages) -- and treat it as an individual record in a database. You can attach whatever words you want to this chunk of text to indicate where it came from, what questions it raises for you, any notes you want to make about it, suggestions, etc. You can change these terms/notes at any time, and you can add, delete or combine terms as you go along, but at some point of working with your texts and FileMaker you have lots of chunks of text, each an excerpt from a particular text file, and about each of which you've noted something of interest, even in cases where -- and this is what's MOST important -- ALL you've noted is that this particular chunk is interesting, for whatever reason. You can then pull up everything you thought was interesting for what it suggested about x, or y, or z from all your different text files, look over everything you thought was interesting, and try to figure out what the hell you were thinking when you noted that in the first place. You can then make some more notes about that and add those to what you already have. For lack of a better word, I tend to call this kind of annotating and wondering and re-annotating, "data analysis." Many qualitative data analysis programs can be helpful in doing this kind of annotation and retrieval of interesting chunks of text. The problem is that many of them also make it TOO easy to skip the annotation, wondering and re-annotation process (the "data analysis") and just plunge ahead with "coding" -- which, as I've noted above, implies that you have already determined the theoretical significance of your data, before you have analyzed it. Annotape: A lot of theorizing has already gone on for many researchers in just getting their data into "text files" in the first place -- another necessary evil of doing social research. But maybe not always necessary. . . If you are working with audio or video tapes, and you'd like to analyze these in the way I've noted above (annotating, wondering, re-annotating, etc.) you might want to look at a program called "Annotape" that lets you write notes on audio files BEFORE you transcribe them -- not just summaries of the whole file, but annotations that are attached to discrete segments of an audio file. Annotape also works like a data base, so after you've listened to the audio and attached your notes to specific sections, you can find every section from every tape you've "annotated" that refers to term/question x, y, z, etc. This allows you to do the annotate-wonder-re-annotate kind of analysis on the audio itself (which always provides a richer data source than a text transcription). If you want to transcribe, you can also use the program as a full-featured software transcriber -- which can be more economically, because instead of just turning everything into text you can focus on segments of audio that you have already determined through your "data analysis" to be really significant. Both FileMaker and Annotape work with images as well as text and are available for both Mac and PC's (while the PC version of Annotape works with audio, text and images, only the Mac version works with video). You can try out Annotape by downloading a demo. FileMaker is widely used in lots of administrative offices, so you might be able to find someone nearby who can show you how it works. Both programs fall short of the automatic "theory building" features of some other dedicated qualitative data analysis programs, but I find those features somewhat problematic -- for the reasons outlined above. Having said all this, there's also lots you can do just by reading and making notes -- in whatever format, media or program you're comfortable with. Jon WagnerWhat Joe says below seems very sensible, but I'll up the ante one notch-- my own recommendation is not to use qualitative software at all. Every program I've seen asks the analyst to make categorizing judgments too soon in the analytic process, and then makes it difficult to change the judgments in subsequent passes through the data--that iterative process (slightly misleadingly called "constant comparison"--it's not constant, it stops when you stop because you've ruled out competing interpretations) is the essential thing. More generally this process of recursive analysis of data, changing your mind as you go along, is called "analytic induction". For my money it's far more useful to read about analytic induction--get clear about what analytic induction is, and how to "do it" by hand on sets of diverse information sources (e. g. field notes, interviews, site documents, contextual demographic information, etc.)-- than it is to try to use the software packages. Of the software I've reviewed NUDIST seems to have the least of this press toward "premature categorization" built into it--but I still think that if you are a single researcher, collecting your own information material, to generate (identify) data from the information sources by hand and then to analyze the data by hand is still the best approach, and that is what I tell the students in my data analysis class. Frederick EricksonOn 9/7/03 10:25 AM, "Joseph Maxwell" <jmaxwell@GMU.EDU> wrote:Kevin: This is a little bit like asking for recommendations on what kind of house to buy for a family of four. I'd need to know a lot more specifics about what kind of data you have, your research questions, and your conceptual framework to even make a guess at what would work best for you. For a detailed strategy for selecting qualitative analysis software, see Computer Programs for Qualitative Data Analysis : A Software Sourcebook, by Eben Weitzman and Matthew B. Miles (Sage); the descriptions of particular programs are somewhat out of date, but the general advice is excellent. For an explanation of the difference between categorizing and contextualizing types of analysis, and the implications of this for selecting software, see my book Qualitative Research Design, An Interactive Approach, pp. 78-81. Joe Maxwell At 10:44 PM -0700 9/3/03, Kevin M Foster wrote:I am looking for recommendations regarding qualitative software. ... more specifically, the use would be for the analysis of gender-specific ethnographic/qualitative data. recommendations? -KevinJon Wagner Professor, School of Education, UC Davis Office: 2397 Academic Surge, UC Davis E-mail: jcwagner@ucdavis.edu PH1: 530-752-5387 PH2: 510-527-5199 FAX: 530-752-5411 Mail: School of Education, UC Davis, 1 Shields Ave., Davis, CA 95616 Image Editor, CONTEXTS http://www.asanet.org/contexts/ ********************************************************
[This page brings together a series of exchange about indexing, coding, computer programs for ethnography and related methodological matters. Most contributors have much experience with the analysis of ethnographic material. Many teach versions of the course where I use this exchange as supplementary reading.
[to preserve the chronological order of these exchanges, one should read this starting from the bottom up]