I am using SmartPLS for my research. In my research, one of my variable's (Dependent Variable) AVE is 0.408 but the composite reliability is good. All other variables show good AVE (above 0.5) and composite reliability (0.6 to 0.9). In such case, can we ignore AVE if the composite reliability is good.
Any suggestions along with the supporting material will be highly appreciated.
Note: My research is exploratory study.
Dear Chaudry,
have a look at the Cronbach's alpha value of your variable: if both CR and α show satisfactory values (CR>0.80; α>0.70), I'd say the variable is reliable (actually, your AVE is approaching 0.50). However, 2 main issues:
1) did you choose already validated variables in previous literature? In this case, have a look at what values they found and compare them with yours; maybe it's a sampling issue.
2) the validity and consistency of your variables could be first assessed by an EFA (exploratory factor analysis): maybe you could think of deleting the items that show low factor loadings to the latent variable; try to run a PCA and see what you got. Next, you should run a CFA (confirmatory factor analysis) to assess the fitting indexes of your constructs.
Have a look at the following James Gaskin's video tutorial on EFA with SPSS: https://youtu.be/VBsuEBsO3U8.
Hope this help,
best luck with your analysis.
Lamberto Zollo Thank you very much. Actually it is Hughes and Morgan (2007) Entrepreneurial Orientation (EO) Scale. I have run EFA on it. I think I have caught the problem. The EFA of EO doesn't produce the dimensions as expected. I have developed my own scale about management functions which I am testing against EO. My function is giving a good reliability, composite reliability and AVE. But EO AVE is 0.408 due to overall Model Fit is not being achieved. Looks like I better change the scale of EO and jump to Lumpkin and Dess (1996) or Miller (1983).
Thanks once again.
- Badges
- Science method