While performing EFA using Principal Axis Factoring with Promax rotation, Osborne, Costello, & Kellow (2008) suggests the communalities above 0.4 is acceptable. This is also suggested by James Gaskin on http://statwiki.kolobkreations.com . However, Child. D (2006) suggests that the value of communlaity below 0.2 should be removed.
In my case, the communalities are as low as 0.3 but inter-item correlation is above 0.3 as suggested by Field. A, (2009). What is the cut-off point for keeping an item based on the communality? Any other literature supporting (Child. D, 2006)?
Hello Chaudry,
There is no fixed level for what degree of communality qualifies a variable as having sufficient common variation to retain in a factor solution. You will, however, find various suggestions, as in the one from Osborne et al. (2008).
The same is true for identifying a variable-factor loading (correlation) as sufficiently high to be "salient." Raymond Cattell gave an example in which a loading as low as .15 was judged salient.
What you will find, however, is that as your threshold for variable communality increases, you will see: (a) fewer variables retained; (b) a "cleaner" solution as regards rotated results matching to Thurstone's idealized "simple structure" result; and (c) possibly/probably fewer factors (as a result of a).
Good luck with your work!
Hi Chaudry Bilal Ahmad Khan
communalities is calculated sum of square factor loadings. Generally, an item factor loading is recommended higher than 0.30 or 0.33 cut value. So if an item load only one factor its communality will be 0.30*0.30 = 0.09. So, an item communality can be 0.30, because of calculation of its value, i think. If factor loandings high and total explanied variance sufficent, it's ok.
Bests
Article Exploratory Factor Analysis for TPACK among Mathematics Teac...
On an abstract theoretical level, communality can be any in 0 -100% range, and won't violate FA model or "quality". But on the practical side, there is at least a couple of points for "large communality is preferable" suggestion. First, obvious, in case we are doing FA for designing tests/questionnaires, we are not interested in items irrelevant to what the instrument is meant to measure, i.e. items with high uniqueness, away from the "packs". Second, less obvious, high communality of items demands a lesser sample size for FA to be able to uncover population factors correctly.
- Badges
- Science topic
- Similar topics
- Psychology