,
10 tweets,
5 min read
Read on Twitter
Pro tip: Don't start a discussion of data set coding 15 minutes before you have to go teach. Here's my attempt at a re-cap:
1/ @ProfPaulPoast starts off by discussing teaching grad students about coding decisions, and the difficult decisions:
1/ @ProfPaulPoast starts off by discussing teaching grad students about coding decisions, and the difficult decisions:
2/ great discussion bt @DenisonBe, @sbmitche, @KSchultz3580, @BridgetCoggins, @dhnexon abt difficulties of coding decisions, and the way that datasets might be (a) a product of their time, and (b) get reified when imported into other large data projects.
3/ @sbmitche crucially highlights the importance of reading codebooks, and @KSchultz3580 notes that there has been a useful move toward providing fuller documentation of coding decisions.
4/ larger discussion about the critical questions raised by data, quant AND qual, and the importance of understanding context of when piece/data was written/collected
5/ resources mentioned throughout the thread:
Internatl systems dataset: researchgate.net/publication/26…
@IdeanSalehyan JPR special issue on conflict data: journals.sagepub.com/doi/pdf/10.117…
Wimmer & Min on empires and nation state: journals.sagepub.com/doi/abs/10.117…
Internatl systems dataset: researchgate.net/publication/26…
@IdeanSalehyan JPR special issue on conflict data: journals.sagepub.com/doi/pdf/10.117…
Wimmer & Min on empires and nation state: journals.sagepub.com/doi/abs/10.117…
6/Reflections: (1) coding is hard work - the scale of info that goes into dataset is massive; (2) context: work cannot escape the time in which it was developed. There is a euro-centric bias in much of the field, which then gets imported to other datasets and case studies...
7/ ...be aware of this, not to discard datasets, but to understand collection process. (3) path dependence: change over time and time series analysis incentivize keeping small potential problems in data. and we want to study change! there's a ripple effect over time...
8/...decisions at time 0 get amplified as more scholars use the evidence and build it into additional datasets.
what to do? Be aware. cultivate relationships with case and area specialists. work together, rather than compete bt quant and qual
what to do? Be aware. cultivate relationships with case and area specialists. work together, rather than compete bt quant and qual
9/ thank the people who took the time and effort to collect datasets. Be humble. we can learn a lot from quant datasets, but need to ensure we're cultivating quantitative and dataset literacy, so that we're aware of promises and limitations in our sources of evidence.
10/ tl;dr My dream: quant AND qual is not a rivalry, but a team.
Thanks for all of the responses and engagement. Enjoyed coming back to this thread after my morning class
Thanks for all of the responses and engagement. Enjoyed coming back to this thread after my morning class
