I use data from 300 questionnaires mostly with likert scale. I am trying to figure out which method fits best. Would it be a contribution to compare results using both?
Dear Meletios,
It has become very common in business studies, particularly in articles published in the Journal of Business Research, to use both QCA and SEM in a single study on the same set of data.
However, the question of what the value added is remains. Causal structures targeted by QCA have a very different architecture than those targeted by SEM. Results are incomparable because the outputs of each method, and their respective interpretations, are completely disparate. The algebraic logic underlying these two methods is incompatible, and no one has so far presented an argument how they could be reconciled (see this article http://cps.sagepub.com/content/49/6/742.abstract).
I know it has become en vogue to use multi-method research, regardless of what methods are combined. But the mixing of QCA and SEM seems to me, at least given the current state of knowledge, to be a case of "Hazardous Hotchpotch" instead of "Enlightened Eclecticism", as Wolf (http://mmr.sagepub.com/content/4/2/144.abstract) has aptly described it. Thus, if I were you, I would only use one of these methods, but try and make sure its implementation is as good as possible.
Best wishes,
Alrik
Hi Alrik,
Thank you very much indeed. I agree with you. Asking is because I ve noticed this kind of mixed methodologies in JBR. The links you provided me with are quite illuminative. Which statistical package do you think is the best for a beginner like me?
BR,
Meletios
@Alrik Thiem
Hi, the replication data on the journal page is no longer available. Could you maybe share that from here? Thanks!
Dear M. Murat Ardag,
Thanks for letting me know about the replication data problem. SAGE seems to have deleted it when they implemented their new websites. We've had this problem a couple of times now already. You find the replication folder on my website at
http://www.alrik-thiem.net/about/ (article entry 16)
Best wishes,
Alrik
http://www.alrik-thiem.net/about/
Dear Meletios,
With respect to SEM: if you have the possibility (several universities do this), outsource the data analysis to your statistics department or your university's statistical consulting services. They are the most competent people. If you do not have this possibility, you could use SPSS Amos, for example, for which you find many tutorials, both in text as well as on platforms such as YouTube. Obviously, there exist other software packages such as STATA (https://www.stata.com/features/structural-equation-modeling/), which you could also use.
With regards to QCA: beginners often chose graphical interface software such as fs/QCA (http://www.socsci.uci.edu/~cragin/fsQCA/software.shtml) and Tosmana (https://www.tosmana.net/). However, there's a major drawback here! These two programs operate on biased model selection by hiding many viable models from their users (http://www.alrik-thiem.net/teaching/). The only currently available software package for QCA that standardly produces unbiased results is my QCApro package (http://www.alrik-thiem.net/software/), which is an extension package for the R environment.
Best wishes,
Alrik
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Applied Statistics
- Survey Design
- Questionnaire