Hi, Several studies have used Cronbach Alpha to test the consistency between the groups of Likert Scale surveys. I have doubts that some of these studies have made a mistake by using this method since the results sound irrational. I wish to know in which cases this coefficient should not be used to test consistency? Can you recommend a reference or more about this topic?
Thank you in advance
Dear Badeel,
This is one of my opinion. I think Cronbach' alpha is a widely used index that reflects the consistency reliability. As a consistency reliability, it assumes that the items of a scale reflect an identical factor. Therefore, we cann't report a single cronbach alpha index if the scale includes several unrelated dimensions.
Dear Badeel Al-Mahdawiy
,There are many myths and misconceptions about Cronbach's alpha:
- Myth #1: The more items the better: An optimum number is 4 to 6
- Myth #2: A score > 0.7 always indicates a scale is valid: It depends on the context and the number of items in the scale
- Myth #3: You always need a sample size of at least 300: It can be used in some circumstances with samples as low as 30
- Myth #4: Cronbach's alpha is the best technique for establishing scale reliability: A better technique is Principal Component Analysis as it is possible to get a high score on Cronbach's alpha with several underlying components
Please see my study guide on scale reliability with small samples for more information.
if you already valid tool (standard tool) then Cronbach Alpha should not be used as a method to test consistency of tool
- Badges
- Science topic