I fixed the references and add some DOIs.
I also fixed line 82 (remove reference to section).
I also tried to polish the tables. We can do better but they are readable. That’s the most important
1 Like
Is it possible to rename NGI Forward to NGI. The tables will be much better
1 Like
Yes, of course.
Good. I fixed TAble 2 (numbers for stacked networks missing) and many small stuff here and there.
Ok, so:
- added the abstract
- tweaked the figures so as to make the color scheme uniform
- resolved most, but not all, of the comments.
To do:
-
@Jan, I am having trouble with this sentence:
In doing so, we highlight the affinity of each of the four techniques with a prominent method of analysis associated in turn with an identifiable school of thought in sociology or anthropology. Our objective is to contribute to the rigor and transparency of the methodological choices of researchers when dealing with large ethnographic corpora.
I am not sure that what you then describe are methods of analysis, as opposed to research interests (for example, “identifying the central concepts in a discourse” or “discovering hegemonic and counter-hegemonic clusters of meaning”). It’s important to get it right, this is the added value of the paper and goes in the abstract.
-
We are still in need of a decision about the node-link diagram figure 5 – @melancon.
-
What is purposive sampling, @Richard, and what do you mean with your comment “I think we should refer to the impact this would have on association breadth”?
-
Read and respond to the remaining comments.
-
Anonymize proper – we’ll do this last.
Now familiarizing myself with the submission website.
Purposive sampling is type of non-probability sampling, the aim of which is to produce a sample that can be logically assumed to be representative of the population. If you want to interview 50 people who are broadly representative of the general population but you find that the first 30 are predominantly women, say, you would use purposive sampling to target potential male respondents, so that in the end you will have roughly half men and half women. (We talked about doing this with our project to find more right-wing participants, who are underrepresented in our interviews.) I wondered whether this would have an impact on association breadth given that the network hasn’t emerged ‘naturally’ but has been shaped by our desire to have a particular mix of respondents.
1 Like
Why only breadth? It would certainly have an impact on the corpus, and so on its network representations. I would file that under “deciding what constitutes data, which is an interpretive choice”. Once the dataset has come together and you have decided to apply one technique, the network will emerge “naturally”, in the sense that people have not been nudged to agree or disagree with each other.
In the case of right-wing informants, of course, they are likely to agree in supporting right-wing leaders, but you do not ask them if they do. Of course, in that case you would get what you put in! Rather, you try to see how they live their lives, and you might well find that they have points of view in common with people on the left.
Almost there.
@Jan and @Richard, could you have one last look at the Related work section? I worry that the exciting discussion of your Mapping network reduction techniques onto methods of analysis in sociology and anthropology section is adequately placed within the debate (I worry not because I doubt of your work, but because I am utterly ignorant of your fields, of course). For now we have:
It is important to maintain full awareness of the implications of applying each technique. In this sense, this work is inscribed in the tradition of scholars who aim to apply systematic visualization techniques, while still retaining sensitivity to informants’ contextual, interactional, and socioculturally specific understandings of concepts \citep{Dressler2005, Hannerz1992, Strathern1996, Burrell2009}.
Is it enough?
I believe this is done, right? – but I didn’t do it myself …
You cite the paper by Beaulieu and Leonelli 2021 to in a sense underline the usefulness of our reflection on the quantitative approaches we use. A 2018 paper by Ollion you might know is in the same vein, although he points more specifically at issues raised by machine learning methods used in the social sciences.
I just thought I should mention it here. I am not sure whether it can back our argument.
Bruno fixed the LaTEX 
I am going through the whole paper, and gosh do I like this stuff 
While reading section “Techniques for network reduction” (that’s about where I’m at now), I got the impression we need to be cautious and make sure the reader grasps all the necessary technical details. See the comments I made in the overleaf doc.
One question came to me while reading: do we really need to describe things in terms of two-mode and one-mode network? I know we all like this way of looking at the data, but I believe we can go away from this perspective. What we need to explain is that we start from annotated contributions, and that we create a network of codes by connecting them the minute they co-occur in a contribution.
We thus create a link between codes A and B also containing an information indexing the author of the contribution. We also need to explain that in doing so, we introduce multiple links between the same two codes, because the same codes can co-occur in more than one contribution – authored by another author or by the same one.
An important step is when this network (which is called “relative” – see my comment in the overleaf doc) content is aggregated in a new network, the one we call the stacked network. I emphasize the use of the term “aggregated”. We should not point at this step as a reduction step (!).
More in the overleaf, and more to come.
1 Like
Good point. This might improve readability. I will read your comments and follow up.
In the latest version, I decided to make this a preliminary step of all four reduction methods. This is because…
… exactly that. We can argue that no information is lost in stacking, and do not use the stacked for analysis. So, I derubricated stacking to a preliminary step, and removed discussion of the nature of the stacked.
Dear All, I am on a sort of family vacation on North Carolina, but I will do my best to do my read through today. There are several places where I want to make minor adjustments/edits. Have a wonderful “pre-new year’s” holiday time!
1 Like
@alberto or anybody, like @Richard I have no access to “toggle on.” And the software does not seem too many editing tools. So, I am marking the text to be removed in boldface and simply inserting the new text. J
1 Like
@alberto and all, as the family is going shopping (an American post-Christmas custom, kind of curious, as Joseph, Mary and Jesus most likely did not do much shopping in Betlehem), I should be able to finish soon-ish. I am writing down the numbers of all paragraphs I am changing/editing and when I am done I will share with you. @Richard - I do not have any good idea about purposive sampling… @alberto - should we try to insert a few sentences that would explain our concept of ethnography? Our data is not generated as notes from participant observation, but rather as transcripts of quasi focus groups or in-depth interviews. What is left of the “classical” ethnographic approach is what we can call an ethnographic sensitivity, that is an attempt to reconstruct a picture of the world from the “native’s point of view.”
@alberto I take it back! What is written in the opening paragraph (number 40) is well-done. I just edited it slightly.
1 Like
@alberto and All, I am done. It is a truly interesting paper. I am unable to assess the value of your technical work, which I am sure is pretty high, but I am very happy that we managed: (1) to stay sensitive to the data collection and (ethnographic) interpretation side of things, and (2) to spot and clearly outline the connection between the four reduction techniques and theories that are (in my mind) associated with each of them. Here is the list of my edits/corrections/additions (please review):
40 – please read the whole paragraph carefully. It is very key. It is well-done I think, but I edited it a bit.
52 – opening sentence considerably reworked
65 – serious re-write
337 – new, rewritten text. Please review.
343 – Richard on purposive sampling
347 – Gramsci not Buttigieg (where did this come from?). AND: it is Laitin alone. No “Laitin DD, Watkins IV JT et al.”
347 – last sentence – rewritten
440 – last sentence: added “fleshed out and”
2 Likes
I’ve read through the article again and looked at Jan’s edits. There are a few phrases which sound a bit odd (school of thought approach; approaches methods of analysis) but I’m not sure if they are intentional or the result of editing. (See below.) I’ve also left a few messages in the margins, the most important of which I’ll repeat here:
15: As I am at UCL only and not Rutgers and UCL, I’ll need a separate affiliation footnote - but I can’t work out how to do it.
29: Do you mean ‘school of thought’ rather than ‘school of thought approach’? As structuralism isn’t a method of analysis but a school of thought, I don’t think this formulation quite works. I’d suggest we remove what is written after the colon. It makes sense when we discuss it in the later section.
333: ‘four major approaches methods of analysis’ is unclear. Four major methods of analysis?
337: ‘School of thought approach’ sounds funny. School of thought?
343: Perhaps we could just make reference to the fact that the choice of purposive versus probability sampling will have an affect on the nature of the network. As in all small-n qualitative research, the sampling procedure is a judgment call, depending on how important the specific make-up of the sample is for a particular project.
I think we’re nearly done! It’s in really good shape.
Google Scholar does not find Gramsci, but only books where the Prison Notebooks are edited by some scholar, in this case Buttigieg. How do I quote the man himself?
This is what GScholar has on Laitin in 1986:
This is confirmed by his own CV. DO you remember the title of the article/book?
This I am not so sure about (see).
Thanks Richard, I will look into it and let you know.