I have a sclae consisting of 9 questions, and all of its questions are answered in the yes/no format. I want to compute the internal consistency of this scale. Is it OK to use Cronbach's alpha in this situation?
Hi Sedat,
This is OK practise, however coding it as you suggested (0 for no and 1 for yes) is recommended, and you should be aware that the dichotomous nature of responses means the assessment of internal reliability will be less robust and sensitive.
Best wishes,
Tom
Sedat,
Yes, you may compute and report Cronbach's alpha. When all variates are dichotomous, coefficient alpha yields the same result as Kuder-Richardson formula 20 estimate (KR-20). The only difference is, KR-20 works only with dichomotous measures, whereas alpha may be used with polytomous measures as well.
Good luck!
This paper helped me a lot in understanding reliability. I think it addresses dichotomous data at least in passing:
Cho, E. (in press). Making Reliability Reliable A Systematic Approach to Reliability Coefficients. Organizational Research Methods.
According to Wikipedia, Kuder-Richardson formula (KR-20) is what you need.
In psychometrics, the Kuder–Richardson Formula 20 (KR-20) first published in 1937[1] is a measure of internal consistency reliability for measures with dichotomous choices. It is analogous to Cronbach's α, except Cronbach's α is also used for non-dichotomous (continuous) measures.[2] It is often claimed that a high KR-20 coefficient (e.g., > 0.90) indicates a homogeneous test.
Dear Sedat,
ViSta "The Visual Statistics System" provides some interesting tools for binary-item analysis (including KR-20, specific data visualization techniques, etc.). The program also includes factor analysis for binary data
Please, find attached a paper describing some of the ViSta's capabilities for item-test analysis and visualization.
This software is freely available at: http://www.uv.es/visualstats/Book/
Hope it helps!!
Ruben
Thank you once again for the replies. So reporting Cronbach's alpha values will be acceptable? Or should I calculate the Cronabch's alpha and report the result as KR-20?
Hi Sedat,
Just report it as Cronbach's alpha. If you use SPSS, the reliability procedure (Analyze/Scale/Reliability) generates alpha by default.
Alpha is the most general way. It works with lots of scales. The one technique developed explicitly for the scale you have is dichotomous items is KR-20.
From the classic work by J. Nunnally "Psychometric Theory" "... coefficient alpha would be the correct measure of reliabality for any type of item, and the special version of that formula, KR-20, would be used with dichotomous items" (p. 226)
Como te responden el Alpha se corresponde a otros tipos de escala, si vos contas con respuestas dicotomicas, si no el KR20. el alfa mide la esfericidad de una escala con varios puntajes.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Coding