Hello everyone
Thanks for the fascinating thread of thought! It seems to me that the issues being raised go well beyond coverage vs. consistency as we are now moving towards scientific vs. statistical inference. To me this suggests that there is some ambiguity with regard to the objectives of fsQCA. What is the overall goal? Is it to transcend the qualitativequantitative divide or to complement either approach? I have been thinking about this question long and hard over the last few years and I reached the following (preliminary) conclusion. If the objective is the former, one needs to be clear that the ontology/epistemology fsQCA is based upon is neither postitivism nor relativism but critical realism. If the objective is to complement existing approaches, then fsQCA could be used, for instance, alongside conventional approaches (e.g., surveys). I have found few studies that follow the former path as the vast majority of studies have taken the default option to consider fsQCA as a technique that is complementary (or even superior) to standard regression analysis. In my work, I am trying to show how fsQCA can fit critical realism underlying ontology and epistemology with its conception of multiple, conjunctural causation. But this requires more intensive studies and more indepth knowledge of the cases at hand (and as a consequence, modest generalisations that, at best, apply to the small number of cases under investigation). I would be very grateful if anyone could point me towards more qualitative uses of fsQCA (perhaps falling within the critical realist paradigm). Thanks in advance for your help.
Federico
________________________________
From: Paolo Dardanelli [mailto:[log in to unmask]]
Sent: Fri 08/06/2012 15:30
To: [log in to unmask]
Subject: Re: QCA consistency
Ok, Wendy. I've looked at Ragin (2008) and seen the rationale for using consistency scores.
Thanks a lot,
Paolo
From: Qualitative Comparative Analysis and Related Methodology Debates [mailto:[log in to unmask]] On Behalf Of Wendy Olsen
Sent: 08 June 2012 15:02
To: [log in to unmask]
Subject: Re: QCA consistency
Hello
Re Necessary cause.
I have not used the alpha or p value tests found in the 2nd half of the fuzzy set social science book 2000 because I was uncomfortable with the logic used there. Notably there is not random sampling, nor indeed any sampling, in the cases in that book. They are countries. Case selection was somewhat arbitrary. To me the logic to use in such situations is not inference from sample to population.
I have suggested ways to use the more recent book dated 2008 to clarify the methods to be used.
If you really need a p value then you will need to think beyond the Qualitative Comparative Method a bit. You would certainly want to be very aware of your sampling scheme. You might think of weighting the data by changing frequencies to adapt it to make it more representative. You also would want to think about the differences between a scientific statement, such as a modal generalisation, versus a statistical inference. For a 'scientific inference' we need more than a p value or alpha level... we need clarity about ontology, concepts, relationships, structure, and causal mechanisms. Good luck
Wendy Olsen
... ... email from ... Wendy Olsen
Senior Lecturer in SocioEconomic Research
Social Statistics
School of Social Sciences
University of Manchester
Manchester
M13 9PL
From: Qualitative Comparative Analysis and Related Methodology Debates [mailto:[log in to unmask]] On Behalf Of Paolo Dardanelli
Sent: 08 June 2012 14:54
To: [log in to unmask]
Subject: Re: slides about QCA including consistency formulas
Dear Wendy  as well as Carsten and Pablo,
Many thanks indeed for your help, much appreciated. I've run the formula with my data and it does indeed give the same result as the software. I'm still a bit puzzled, though, regarding application and interpretation. Ragin (2003: 1924) discusses the test of necessity applied to his data on IMF protests in terms of the proportion of cases where Xi?Yi, not in terms of the consistency results given by the formula. The table on p. 194 (table 8.3) headed 'Results of FuzzySet Analysis of Necessity' reports proportion results. His table in Ragin (2000: 114) also seems to suggest using proportions.
My key question is thus: when assessing necessity, should we look at the proportion or at consistency? In relation to my own data, the choice between the two leads to rather different substantive claims:
a) looking at consistency (0.932) I understand I'd be able to claim that the condition is 'almost always necessary' for the outcome as the figure is above the commonly accepted threshold of .80
b) looking at the proportion (21/24 or .875) and using Ragin's (2000: 114) table, though, I'd only be able to claim that it is 'usually necessary' as it only passes the .65 at ?=.05 test...
Which of the two would be the correct interpretation?
Many thanks once again!
Have a good weekend,
Paolo
Ragin, C. (2003), 'FuzzySet Analysis of Necessary Conditions', in G. Goertz and H. Starr (eds), Necessary Conditions  Theory, Methodology, and Applications. Lanham, Md, USA: Rowman & Littlefield, pp. 17996.
Ragin, C. (2000), FuzzySet Social Science. Chicago, Il, USA: University of Chicago Press
From: Qualitative Comparative Analysis and Related Methodology Debates [mailto:[log in to unmask]] On Behalf Of Wendy Olsen
Sent: 08 June 2012 12:06
To: [log in to unmask]
Subject: slides about QCA including consistency formulas
Dear all,
Re PAOLO's enquiry about Nconsistency.
Charles Ragin has published the formulas for Consistency of sufficiency in several places including his book Redesigning Social Enquiry. I enclose my own slides which present these formula in the context of a lengthy training session (taking up around half a day) where we also did practical exercises.
The 'Nconsistency' or Consistency for necessity is on the Annex comprising the last 2 slides. Answering Paolo's question explicitly, this reads:
* Consistency (Xi ? Yi) = ?(min(Xi,Yi)) / ?(Yi).
You have to take the minimum of the fuzzy sets to get their intersection first, then sum across the rows, giving you the Numerator.
The denominator is easier; it is the sum of the fuzzy sets in the rows of the column you mean as Outcome, i.e. Y.
Thanks to Charles for his great clarity about this matter which is not covered explicitly in the Sage volume on Fuzzy Sets by Smithson & Verkuilen. They on the on the other hand mention the 'inclusion ratio' and this lacks the denominator. Thus in Smithson and Verkuilen's book one does not find any distinction of sufficiency from necessary cause. This is confusing because Sage offer their series of books as landmarks. In the Longest & Vaisey STATA journal discussion, Consistency of Sufficiency is called SConsistency and this is the following ratio:
* Consistency (Xi less than or equal Yi) = ?(min(Xi,Yi)) / ?(Xi).
Best wishes I hope this clarifies yesterday's emails which, on reflection, I thought had too many acronyms.
Yours warmly
Wendy Olsen
... ... email from ... Wendy Olsen
Senior Lecturer in SocioEconomic Research
Social Statistics
School of Social Sciences
University of Manchester
Manchester
M13 9PL
**
