2014_11 van Exel

Talk at Jacobs University

26.11.2014, Reimar Lüst Hall, several people present (Olaf, Christopher)

On the number of statements, van Exel advcoates using not soo many (not more than 40)

the information is between the statements; if you have a lot of statements, there is not a lot of room between the statements van Exel

Get the latest paper by van Exel; he did a q-study on different conceptions owe f the good life, including Stiglitz, Nussbaum, OECD, etc. (great!)

Notice that I am a little bit unhappy with the standard for developing statements by van Exel; the "pieces of carpet" analogy is, in a way, much more exacting – and maybe less politicized.

Finding very different people for my study was obviously hard; people were not so interested.

Snowballing might be an interesting idea for the future of CiviCon, too: ask subscribed participants to suggest/invite other participants that they will disagree on the topic with.

I don't have a person or viewpoint that woud be more important to rotate to.

van Exel wants a one-stop-shop for Q, maybe a new PQMethod, including online administration; this should be a job for

Meeting over Dinner

Present:

• Peter Holtz
• Christopher Cohrs
• Melis Ulug
• Job van Exel

Issues:

• What is Strukturlegetechnik
• Look into Latent Class Analysis
• Job van Exel recommends "principal axis" as the way to go

Meeting in the morning

Criteria for number of factors to be extracted

• use nloads > 2, eigenvalue
• use Use Horn's Parallel analysis
• use the van Exel method for correlation

Rotation & Extraction Method

• judgmental would be nice, but can maybe be done later
• there are a lot of new methods being discussed now, including principal axis analysis that I should probably look into

Before-After Effect

There are several ways to look at the treatment effect, in rough order of adequateness for my study:

1. compare factor loadings: look at the dimensionality, and loadings of factors befre and after; this would be the broadest kind of result, ideally that the factor space becomes lower dimensional
2. pooling of data (including before and after sorts in one dataset for analysis): may show different before/after factors, as well as the change of people
3. 2nd order analysis (make factors out of factors): will be limited because of max. 8 cases (factors), but would be very interesting: these might be the misunderstandings
4. factors of deltas compute the delta in score for each item (say, item moves from -5 to 2 for a participant), then compute factors of these movements: this may reveal the "misunderstandings", and/or the kinds of learning processes that have happened.
5. spiking of data; place (some of the) "before" factor arrays into the "after" dataset as cases; then do judgmental rotation: this is not a very common method, but may reveal how things have changed from the vantage point of "before".

Visualization

... is key to make quick sense of the data, especially in the form of factor arrays.

• The color of factors might be somehow scaled/chosen based on the cross-factor correlations.
• Include significance of loading of people on factors in the visualization?

Other interesting stuff

• Look into qblock method – what the hell is that?
• Look at browns factor index approach; people will assign or be confounded to a factor
• Look into latent class analysis
• get back to christopher cohrs,
• figure out political psychology interests of christopher cohrs
• get back to the q guy from berlin, talk to him
• next q conference is in ancona (?) in the next fall.
• Look at the deadline for the q conference in ancona.
• Get in touch with amanda, former editor of OS, who is now in NZ; job suggests to get in touch if I want to spend some time in NZ.
• Comment on the item complexity.
• Have other people sort the qsort as, say, a hyperliberal or something and then use that as a target rotation.
• look at the status quo; do they like

Interesting journals

(actually these are low-impact, niches, but still)

• operant subjectivity
• quality and quantity