Hello, I'm struggling to get this right. In my study, I have used a Teacher Feedback scale (4 subscales) to ask teachers about the frequency at which they produce different types of statements towards individual children in their classrooms. I want to report the internal consistency of the instrument since I made some light adjustments to the original scale.
My data matrix is: children sorted in rows, items in columns (items repeated as many times as teachers). How can I calculate Cronbach's Alpha?
If I follow the rule of one alpha per application, since each teacher rated their feedback in relation to each child, I would obtain: 3 teachers x 4 subscales x 9 children (in one classroom); and 2 teachers x 4 subscales x 10 children (in the other classroom).
Please help! Thank you.
An important question might be whether or not you expect the children to act or behave the same for each teacher. For internal consistency, do you expect the child's interaction with the teachers to be consistent, or might there be personality differences that result in differences in teacher response concerning individual students. There are online tools: http://www.wessa.net/rwasp_cronbach.wasp or http://www.real-statistics.com/reliability/cronbachs-alpha/ that can be used. Another measure is inter-rater agreement (or KAPPA): https://www.medcalc.org/manual/kappa.php
Alvaro, in addition to Thomas' suggestions it seems that inter-rater agreement would be the best choice. Cronbach's alpha should seldom be used because of the very strict assumptions, such as unidimensionality, tau-eqivalence, and uncorrelated error terms. These assumptions should be checked, and they are almost always violated.
Regards, Karin
I am also not sure I unerstand the setup of the study. As far as I know, Cronbach alpha requires independence of measurements and if each of your participants rates many teachers, this requirement is not met. My idea would be to use multi-level analysis (teachers nested in students), and then the models that are tested do bring a reliability coefficient.
- Badges
- Science topic
- Similar topics
- Psychology
- Developmental Psychology