I have a question regarding Cronbach Alpha, and it's cutoff criteria.
Any suggestions are very much appreciated!
I did a factor analysis on 27 items and extracted five factors
(Cronbach alpha ranges from .47 to .68). The summary of findings is
Factor A: 5 items, Crobach alpha=3D0.68
Factor B: 4 items, Cronbach alpha=3D0.6
Factor C: 2 Items, Cronbach alpha=3D0.54 (Pearson corr=3D.371)
Factor D: 2 items, Cronbach alpha=3D0.52 (Pearson corr=3D.355)
Factor E: 2 items, Cronbach alpha=3D0.47 (Pearson corr=3D.326)
My committee asked me to only include the reliable variables (Factor A
& B) in a regression analysis. However, I=92ve seen many published
studies including factors with similar alpha coefficients in the
regression analysis in my field. Plus, factor C, D, and E do have
moderate correlations for each pair of variables, and they all only
have two items. I feel it is still valuable to include all five
factors in the regression analysis, since they are supported by the
theory. I prefer to including at least the first four factors in the
regression analysis (the results are also more interesting), but my
committee don't support the idea.
1. Is it appropriate to use the formula of Spearman-Brown split-half
reliability coefficien to argue that if the factor have 4 items, its
alpha is close to .7?
2. Is there any other reliability measurement that I may use? Theta or
3. Any other suggestions about how I can make the argument are very
Thanks so much for your input on this.