How cronbach alpha coefficient can be manipulated in terms of the data changes? Let's say I have 10 respondent and 3 questions (Scales) with 1-7 likert type scale from Strongly Disagree to Strongly Agree.
what should the answers approximately be to make cronbach alpha coefficient go up more than 0.8?
By asking this question, I want to get the core idea around cronbach alpha. In a practice, how is it impacted by the respondents answers. That is what I want to understand, please advice.
Dear Levan,
I'm afraid I can't answer your question directly as I have never studied the maths behind this statistic. However I would like to make a few points:
- A sample size of 10 is way too small for a trustworthy result. Textbooks generally recommend a minimum sample size of 300-400, but simulation studies have indicated that results can be trusted for sample sizes as low as 30 provided that the leading eigenvalue of a Principal Component Analysis is sufficiently large.
- 3 items is the absolute minimum for a scale - if you are designing a scale from scratch I would recommend starting with double the number of items you want to end up with
- Cronbach's alpha values increase as the number of items in a scale increases, so it is worth considering the principle of parsimony when designing scales
- The interpretation of an acceptable Cronbach's alpha value will also depend upon the context
I have written a brief guide on the small sample size issue here: https://www.researchgate.net/publication/280936182_Advice_on_Reliability_Analysis_with_Small_Samples
Technical Report Advice on Reliability Analysis with Small Samples
- Badges
- Science topic
- Similar topics
- Methodology
- Research Methodology
- Research Papers