We are building an almost "standardized" questionnaire according to Ellen G. White on the 8 natural remedies. When doing the factorial analysis, we found 4 dimensions with Alpha of Cronbach above 0.60, but 4 of them below 0.60. We do not want to destroy the issues. What should we do?
I think the question is a bit problematic. The values you've provided imply that what you did was not a Factor analysis, it is a reliability test to internal consistency using Cronbach's alpha. The recommended values are 70-90. Alphas below 70 would raise concerns about scale inconsistency. See examples below on how to validate measures and feel free to inbox me if you have any questions. Thanks, Ali
Article When empathy hurts: Modelling university students’ word of m...
Article Exploring Personal Experience of Wartime Crisis Effects on J...
Article A Confirmatory Factor Analysis for SERVPERF Instrument based...
Article Linking information motivation to attitudes towards Web advertising
Deleted research item The research item mentioned here has been deleted
Conference Paper Validating a Plagiarism Scale: A Middle Eastern Context
Hi Gina,
in 90% of the cases of a low alpha, the reason is simply heterogeneity. Find out how the latent structure is and calculate alpha (or better: omega) for each specific dimension.
Best,
Holger
I think the question is a bit problematic. The values you've provided imply that what you did was not a Factor analysis, it is a reliability test to internal consistency using Cronbach's alpha. The recommended values are 70-90. Alphas below 70 would raise concerns about scale inconsistency. See examples below on how to validate measures and feel free to inbox me if you have any questions. Thanks, Ali
Article When empathy hurts: Modelling university students’ word of m...
Article Exploring Personal Experience of Wartime Crisis Effects on J...
Article A Confirmatory Factor Analysis for SERVPERF Instrument based...
Article Linking information motivation to attitudes towards Web advertising
Deleted research item The research item mentioned here has been deleted
Conference Paper Validating a Plagiarism Scale: A Middle Eastern Context
Well - I am sorry to say this, because it's always bad to hear that, but low Alpha means weakness of the created tool - nothing more, nothing less. Alpha values below .60 are practically very low values for a reliable measuring tool/questionaire. You should use factor analysis - because I suspect that the tool on the basis of which you created your own questionnaire - contained similar errors (which unfortunately is a common case).
An alternative method is to use in SPSS for Alpha calculations an option which indicates how much the Alpha value will be raised after excluding the given item - and then exclude items until Alpha value increases.
Despite this, it should first used correctly (which is quite difficult) factor analysis and, for example, with the use of an anti-imaging matrix you can exclude weak positions (Field, 2009 - I recommend this book at all).
Correctly performed factor analysis should solve this problem, but at 90% it will end with a discrepancy with the theoretical construct. Well - that's the way it is, do not worry. Good luck!
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Applied Statistics
- Survey Design
- Questionnaire