After I conduct a validity test on my Lisrel, the result comes out. Some of my indicators showed 0.4 score on the standardized loading factor score. I heard that there are some experts that still allowed 0.4 score on the SLF. Is there any journal article or any reference that could help me defending my argument about the SLF score? Thank you :)
Dear Alia,
I don't know what LISREL is but a well cited reference to loading scores greater than 0.4 in factor analysis is:
Guadagnoli, E., & Velicer, W. F. (1988). Relation to sample size to the stability of component patterns. Psychological bulletin, 103(2), 265.
Hello Alia,
Regardless of the software package used (LISREL, R, SPSS, SAS, Minitab, Factor, etc.) in the context of factor analysis, the user does have to make choices.
The choice of how strong a loading need be in order to be considered salient is one that is commonly discussed in texts or chapters on factor analytic methods. Richard Gorsuch's 1983 text, Factor analysis (2nd ed.), Hillsdale, NJ: Lawrence Erlbaum, does a good job of discussing different tactics. Raymond Cattell, a significant contributor to the FA literature, gave an example of using loadings as low as .15 as salient (though these are clearly in the minority!).
Other than using (a) statistical significance (difference from zero); or (b) Kaiser's RMS (square root of average of squared loadings for a factor) as a criterion for variable-factor salience, the values chosen are arbitrary (but not thoughtlessly chosen, one hopes!). Absolute values such as .30, .40, and .50 are pretty commonly seen.
In practice, stronger loadings make it easier to interpret what a factor is all about, and enhance the estimates of internal consistency reliability. From these perspectives, higher is better.
Good luck with your work!
Dear Alia...
Factor loadings less than 0.5 are means that violation of the assumption for convergent validity based on average variance extracted.
If the content validity of scale doesn't get violation, you should remove these items from scale.
Regards,
Fornell, C., & Larcker, D. F. (1981). Evaluating structural equation models with unobservable variables and measurement error. Journal of marketing research, 39-50.
- Badges
- Science topic
- Similar topics
- Journal Articles