Use cronbach alpha to determine whether a factor is reliable?
YT
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
presented below.
Factor A: 5 items, Crobach alpha=0.68
Factor B: 4 items, Cronbach alpha=0.6
Factor C: 2 Items, Cronbach alpha=0.54 (Pearson corr=.371)
Factor D: 2 items, Cronbach alpha=0.52 (Pearson corr=.355)
Factor E: 2 items, Cronbach alpha=0.47 (Pearson corr=.326)
My committee asked me to only include the reliable variables (Factor A
& B) in a regression analysis. However, I’ve 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.
Questions:
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
rho?
3. Any other suggestions about how I can make the argument are very
much appreciated.
Thanks so much for your input on this.
YT