soztag / worktypes Goto Github PK

waiting for https://stackoverflow.com/questions/47714390/shortest-unique-substrings-of-a-vector

this is one of the ideas from the Oberseminr Meeting on May 30

at least ranking (not rating) info on activities, preferably a full time budget allocated to activities.

Only this kind of data would allow proper Q-type analysis, where we look for similar people across activities.

@sabpf the idea is that we try a Q-approach (with people as people-variables) with the BIBB/BAuA data (selected variables on workplace characteristics), and see whether we can find clusters.

Methodological interest would be to try the Q-approach, but with a different algorithm (not PCA; but perhaps Multiple-Correspondence or some other clustering approach).

Substantive interest would be to see whether we can actually find a more nuanced typology of workplaces, including an appreciation of creativity in supposedly "manual" jobs etc.

Moved from soztag/soztag.github.io#6
/cc @maxheld83 @sabpf

this would make it much easier to actually answer the question of what trend there might be.

see https://stats.stackexchange.com/questions/215404/is-there-factor-analysis-or-pca-for-ordinal-or-binary-data# for a start

some of the Marxist activity types (reproduction of use values, such as raising a child) frequently happen outside of the workplace, while the production of exchanges values, by definition, always happens in the context of the formal economy.

We might get around this by simply limiting our research questions to changes in the workplace, but this causes some problems (see full text)

instead of running a fancy dim reduction to empirically find the hypothesised Marxist types of jobs, we could also just assume which of the activities (e.g. "advertising, marketing, public relations") map onto which Marxist category, and just go with that (for trend or sectoral analyses).

This seems obviously defensible for some:

"Advertising, marketing, public relations"
(adv / F309)

Seems to be clearly in the production of exchange value quadrant, while

"Nursing, caring, healing"
(care/F316)

seems to be in the reproduction of use value quadrant, etc.

Once we've built the index, we might still try and check how well it fits with actual jobs, or how people work together.

because we cannot match individuals to check for a change in the share of activities (as would be interesting for a critique), we can, at best, try to match sectors or jobs, and compare the relative composition of activities therein.

This could however be difficult because:

• sector/job (etc.) definitions (as per BIBB/BAuA standards) and their reality probably changed over the course of the waves
• this is, technically speaking, an ecological fallacy.

related to #8

maybe just procrustes rotation or sth like that

also an idea from the oberseminar on may 30