Any thoughts on what is a good proportion of populated truth table rows in QCA? I did a QCA study once with 109 cases and 5 conditions (32 rows). After setting a frequency threshold of 2, I was left with 12 populates rows, containing 80 per cent of the cases. And I got a solution that was perfectly interpretable. But perhaps I was a bit lucky, because I would consider having just one third of the truth table rows populated to be the bear minimum - even if that includes 80 per cent of the cases. I would argue that below 33 per cent populated truth table rows, your solution becomes a function of your logical remainders rather than of the actual cross-case patterns. I would aim to get >50 per cent populated truth table rows. However, there are no guidelines for the minimum number of truth table rows in QCA, as far as I know. So I would much appreciate your sharing your thoughts and suggestions. Many thanks, Roel.
Dear Roel,
good question. But why would you want to care about how many rows are populated? Maybe the causal structure behind your data is such that some rows simply won't be populated. For example, when moving to more complex causal structures than QCA can detect, such as common-cause structures, it will be impossible that some rows are populated. Such structures you can only detect with CNA (Baumgartner and Ambühl 2020) or CORA (Thiem et al. 2022) in any case, but if such a structure is behind your data, it will not help if you require that QCA needs a particular share of populated rows.
In addition, we often don't have as many data as we want, for whatever reason. But even with QCA is it then still possible to detect parts of the causal structure at work behind the data. On this, Michael Baumgartner and I have carried out an enormous amount of simulations in Baumgartner and Thiem (2020, test III pp.296-300). With as few as two populated rows, you can still recover parts of the causal structure behind your data (of course, given the usual provisos of no omitted variables bias, no measurement errors etc.). In this sense, setting some arbitrary threshold for populated rows in QCA will deprive you of many possibilities to carry out valuable analyses.
Best wishes
Alrik
References
- Baumgartner, Michael, and Alrik Thiem. 2020. "Often trusted but never (properly) tested: Evaluating Qualitative Comparative Analysis." Sociological Methods & Research 49 (2):279-311.
- Baumgartner, Michael, and Mathias Ambühl. 2020. "Causal modeling with multi-value and fuzzy-set Coincidence Analysis." Political Science Research and Methods 8 (3):526-42.
- Thiem, Alrik, Lusine Mkrtchyan, and Zuzana Sebechlebská. 2022. "Combinational Regularity Analysis (CORA) — a new method for uncovering complex causation in medical and health research." BMC Medical Research Methodology 22 (1):333.
- Badges
- Science topic
- Similar topics
- Chemistry, Organic
- Organic