tag:blogger.com,1999:blog-4594832939334410220.post951362295644539515..comments2024-02-12T06:23:51.153-06:00Comments on Deeply Trivial: Statistics Sunday: Cronbach's alphaUnknownnoreply@blogger.comBlogger3125tag:blogger.com,1999:blog-4594832939334410220.post-13293430440806470832018-05-02T09:01:41.464-05:002018-05-02T09:01:41.464-05:00Thanks for your detailed answer, Martin! I'll ...Thanks for your detailed answer, Martin! I'll admit - omega is something I was only exposed to recently and don't know a lot about. My grad school training was predominantly classical test theory, with some reference to alpha but mostly validity and threats to validity. My post-doc is in psychometrics, where I learned Rasch and item response theory. I almost exclusively use Rasch these days, which uses a different measure of reliability. I'm planning to write posts about Rasch concepts in the future. In my previous job, where some of our exams were developed with classical test theory, we mostly used split-half reliability and sometimes alpha. Thanks again for sharing your thoughts and for stopping by!Sarahttps://www.blogger.com/profile/13213593768515404983noreply@blogger.comtag:blogger.com,1999:blog-4594832939334410220.post-67402410293369418552018-05-01T09:17:07.536-05:002018-05-01T09:17:07.536-05:00There are theoretical assumptions that need to be ...There are theoretical assumptions that need to be met when using this measure of reliability, and in most cases Omega is probably better measure. I don't think it's suitable to discuss this topic unless we also discuss concept of measurement models.<br /><br />Specifically, it is crucial to understand that when it's said that Alpha relies on classical test theory it's actually a specific model called tau-equivalence measurement model. This model has among other statistic assumptions an assumption of tau equivalence, or that every item could serve as an test of specific latent variable. With alpha it's necessary to check if every item has a similar variance and standard deviation; if not we may have broken the assumptions of the measurement model behind alpha...please don't use alpha coefficient as a "standard". Martin Cadekhttp://www.martincadek.comnoreply@blogger.comtag:blogger.com,1999:blog-4594832939334410220.post-59777715061600304442018-04-01T14:15:53.605-05:002018-04-01T14:15:53.605-05:00Coefficient Alpha is one of the most misunderstood...Coefficient Alpha is one of the most misunderstood and misused coefficients in psychology. Klaas Sijtsma wrote an excellent paper about it in 2009:<br /><br />Sijtsma, K. (2009). On the use, the misuse, and the very limited usefulness of Cronbach’s alpha. Psychometrika, 74(1), 107. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2792363/. <br /><br />For further reading, Jessica Flake and I put together a section on reliability in our reading list "Measurement Matters": <br /><br />https://docs.google.com/document/d/11jyoXtO0m2lUywpC04KjLvI5QcBUY4YtwEvw6cg2cMs/Eiko Friedhttp://www.eiko-fried.comnoreply@blogger.com