Wishlist for Tulip visualization

Wellbeing in Europe Workspace
21

Dear all, based on the second round of our discussion with @jitka.kralova and @SZdenek, we would like to kindly ask @alberto for the following sets of (potential) ego networks overlaps (accordingly to what you explained in comments Mania`s networks) and in one case one extra ego-network.

Political disillusionment
a) distrust, DDeliteestrangement, SIincompetence
b) DDpolcor, Eeconcor
c) grassroots movements, protest voting, civil disobedience, protests

Subjectivity and affect
a) anti-candidate voting
b) resilience, empathy, hope, nationalism
c) anger, alienation, anxiety, fear, hate, hopelessness

History romanticization, declinism
a) invoking history, national history
b) communism, state socialism, victims of post-socialist transformation
c) catching up, euroscepticism, neocolonialism

3 Likes
22

Alright, @alberto
I have looked through the visualization and that’s what I would like to request from you:

  1. Versions of ego networks without gender, with colors representing separate ego networks, and if possible, for joint nodes between ego networks to be in a different color, and with all labels visible, no label overlapping. This request is for the following visualizations:
    1.1. Vis : need for simplicity ego network, tolerance ego network, general freedom ego network, *imagining the future" ego network with cooccurrence reduction 3 and cluster detection,
    1.2. Vis Ego networks for codes abortion , nationalism , and Catholic Church visualized using community detection algorithm (or other algorithm showcasing clusters), no cooccurrence reduction.
    1.3. lockdown ego network and HSlowqual ego network with cluster detection algorithm, no occurrence reduction.

  2. Revisted ego network visualization, but without cooccurance reduction. Even if codes that are this way visible were mentioned just once or twice, they may form along with other codes a certain narrative, that altogether makes sense (like, two various codes mentioned once in that context, but both represent negative emotions). This request speaks to those visualisations:
    2.1 Vis hate ego network crossed with Law and Justice ego network, no reduction, one version with gender showing, one without gender but with coloring as requested above.
    2.2 Vis welfare state node ego, with gender, no reduction.
    2.3 Vis frustration code ego network, with gender, no reduction.

Could we have dual ego network visualisations in each case more polarised in visualisation, to have the nodes in common in the middle?

Thank you!!

23

You can do this kind of exploration yourself, with GraphRyder. For example:

image

This is the result of

  1. using the search box to find the code you want.
  2. Using the slider to select the minimum co-occurrence value (Graphryder uses a variant of d)
  3. Hovering on a node to higlight its ego network and “turn off” the rest of the graph.
  4. Taking a screenshot.

I recommend you save the URL when you do that. Graphryder stores your visualization into long URLs. For example, this one is http://server-2021.edgeryders.eu/dashboard/edgeryders/ethno-rebelpop-polska-interviews?m=cn&cn.weightFilter=3&cn.labelDensity=0.55&cn.labelThreshold=11&sc.code=x%C2%9CK%C3%8EOI%C2%8D7321%03%00%0F%C3%BD%02%C3%8D. Then, if you click on the link, you go back to the same visualization (you still need to hover over the code that you want the ego network of).

The exceptions are 1.2 and 2.2. These I will do, though there is a queue. :slight_smile:

24

Wouldn’t the goal be to have all the visualizations looking the same?

25

On a side note, would you have time to meet with me, @Jan, and @Wojt on some live Tulip network analysis? This way we could perhaps speed up the process without the need to write down detailed instructions, and see the outcomes of different visualizations right away.

26

Between you all, you requested some 80 visualizations. We are not going to put them all in the report (which, IMHO, should be 25 pages, and last time I looked the draft was already at 85), without visualizations. Most of this stuff is just preparatory material, it does not matter what it looks like.

Hmm… it’s getting a bit tricky now for me. I am completely swamped on Tuesday and Wednesday. I guess my only window would be Monday before 18.00.

1 Like
27

@alberto – when I was designing the TOC / layout of the report, you had encouraged me to use Amelia’s Next Gen work report as a model, and that one is 126 pages. Not that I aimed to emulate that length specifically, but that is what I had in the back of my mind as acceptable length.

But yes, of course, all the visualizations are not going to go into the report. In fact, today during the meeting we agreed on five conceptual “pillars” that each of the teams will use, and that will be our comparative analysis basis. They are:

  • trust/distrust
  • incompetent state
  • polarization/fragmentation
  • retraditionalization
  • affect

Each group is going to come up with proxy codes that correspond to each of these. We hope that those codes will produce five visualizations per group that will go into the report. So based on this, I am imagining the report will have a total of 15 visualizations. Apologies to add more to the “order menu” – but this is iterative, and coming out of our discussion of the analysis of existing visualizations.

How does all that sound?

1 Like
28

@alberto

For the German corpus, could we request the following visualisations:

Affect
Anger + Emotionally difficult + Unfair

Retraditionalisation
Retraditionalisation + GENinequality + Patriarchy + Childcare

Polarisation
Polarisation + Anti-vaxxers + Vaccinated + Unvaccinated

Trust/distrust
Distrust + Government + Societal control

’Incompetent state’ doesn’t really apply to Germany. It’s Germany.

Could you see if there is a co-occurrence between ‘Anger’ and ‘Distrust’?

Mille grazie!

29

It’s not forbidden: if that’s the length we need, that’s the length we need. But I do worry about readability. The ultimate line of defense is the executive summary.

I expected iterative. But, because of this, we’d better be economical, falling back on Graphryder whenever possible. For the same reason, if you already have decided on the scheme of 3 languages x 5 pillars, then I’d skip the previous phase and just move to producing the (draft) final visualizations.

Even though I confess that I do not yet know how to map your idea of pillars to mathematical objects in the graph. The graph sees codes and their co-occurrences: if you want to tell a story about trust/distrust, the elements of this story need to be codes and their edges.

30

Right – so we will have “proxy” codes for all pillars (like @Richard just posted upthread)

2 Likes
31

That is a good idea in my opinion

32

With what we are doing right now - the visualizations we are requesting - we are doing exactly that, looking for final visualizations that correspond with our qual analysis somehow, though still bring in a novel perspective (or not, confirm what we found, which is equally great).
Given the high number of requested visualizations, I think it would be great to have the live-vis session with Tulip on Monday, if you are still available before 18. What about we meet after the lunch time, 14 pm CET? This way we could quickly see what visualizations work and skip the back and forth we have gotten ourselves into.

Not sure what time it will be in the US, we have winter time change in our zone tonight.

33

Let’s stop for a moment and consider the stories you want to tell with this data. I want to do that, because this influences the choice of a visualization. For example:

I understand this as a visualization of these three codes, plus their neighbours, plus the edges connecting them. This shows which other codes these the codes in affect are associated to. Is this roughly correct?

@Jirka_Kocian, I see that you have been using a different scheme:

Is this superseded by Nica and Jan’s five pillars scheme?

34

I am starting from @Richard’s requests, as they seem the most likely to be in line with what Nica and Jan want to do, and the least likely to change again.

The highest value of d for which all three codes are still visible in the graph is 5.

image
image1375×897 9.57 KB

Lowering d to 4 gives this:

anger and distrust are not co-occurring at this level of depth. They co-occur only once, in the interview to a female participant.

Note that the co-occurrence between unfair and sharing parental duties has breadth 1, which means only one informant supports it. The others have breadth 2 to 4.

These codes are quite tightly clustered together. All four appear in the graph until d >= 12, where they are connected to one another and to impact of COVID-19 (not shown). For a richer visualization, here’s what happens at d >= 6:

image
image1375×897 18.8 KB

Conceptually, it’s the same. These codes co-occur each other, and three of them co-occur with impact of COVID-19. Two also co-occur with SAsupdef. Also, this is a very female debate… but then this is true of the German corpus in general. The edges involving acceptance and sharing parental duties are supported by only one informant each. The others go from 2 to 14 (! Mr. 14 would be the edge between GENinequality and impact of COVID-19.

These codes are again quite robust, and all three appear up to d >= 9. At d >= 6 it looks like this:

image
image1375×897 28.2 KB

Again, several edges have breadth = 1. If we exclude them, we also drop from the graph the following codes: discrimination, government response, infection rate, lack of vaccinations, civil rights, injustice, and SIbureaucratization.

Here there is a bit of a problem, because the corpus has two codes that look like they should be the same.

Ignoring for the moment mistrust, here is what I get at d >= 6:

image
image1375×897 13 KB

Six of these edges are supported only by a single informant. They involve education system, pandemic, children, children development, respirators/face masks and anti-COVID measures.

I can invalidate this statement with two words: Deutsche Bahn. :laughing:

1 Like
35

The 5 pillar model is what we came up with on Friday, after the Czech team submitted those requests – so I am requesting that @Jirka_Kocian and @jitka.kralova parlay those into the "5 pillars/3 codes"format. Meaning yes, the 5 pillars / 3 codes / 3 languages supersedes previous requests. Having said that, I think the Czech “wishlist” is pretty close thematically to the “pillars” which is good!

1 Like
36

What about the Polish one?

37

@Jan and @Maniamana will need to parlay their latest set of requests into the 5 pillars/3 codes model – are you all meeting on Monday for Tulip-ing?

38

I am not a fan of doing this stuff in real time. I would much rather get the wishlist and work on it. It would be a super long meeting too. It tool me most of the afternoon to do the five visualizations for Richard. But Monday is indeed the last good day for me to do it, so I would be grateful to @jan and @Maniamana for a list ASAP, and I can deliver the viz by the end of Monday.

39

Ok – could the Polish and the Czech groups provide the updates lists for Alberto to do on Monday, then?

@Jan and I have drafted a section explaining the “four principles” we discussed in the meeting on Friday that immediately precedes and will help frame the three ethnographic “country” sections which will contain these visualizations.

40

Yes, please, and Monday morning if that’s at all possible.