Hi everyone
I have a question. For factor analysis in spss, Can one item still be an independent factor in factor analysis or just drop it? It has a high loading. In addition, what if cronbach alpha is low for a factor there are only two items?
Hello Shuyi,
Can one variable be a factor? In theory, yes, but, common factors are (by tradition) thought of as manifesting themselves by multiple observed variables, not just a "singleton." As well, if you were to have a factor solution for which one variable affiliated with one and only one factor (and no others did in addition), then it really isn't common variation in the big picture sense.
So, most folks would reject a single variable factor as representing exclusively variance specific to that variable.
Cronbach's alpha can be low for two-item "measures" unless the correlation between the items was very high. More items = more information = more dependable estimates of the trait/characteristic of interest.
Good luck with your work!
Hello Shuyi,
David has given you some good advice above. Factors (or variables) with only one or two items are a bit "lightweight" in terms of likely validity. Furthermore, Cronbach's alphas are highly dependent on the number of items being analysed, so a couple of items are likely to have a lowish alpha despite being quite highly related to each other, and a large number of items is likely to yield a high alpha despite many of those items having a very small (or some even inverse) relationship with each other.
I suspect the main problem you face is number of appropriate items for what you want to measure. Perhaps, having conducted your factor analyses, you now have a clearer idea about which "domains" you might like to try to create some more (and really good) items for.
Robt.
I'm happy for David (or others) to provide some advice, even if it's contradictory to what I recommend, but if your two variables have formed a separate and strong factor they could well be used as a combined, but limited indicator of whatever they are tapping. Your single variable could be used as a more limited indicator for whatever it is tapping.
Of course, without seeing your actual results, it's difficult to provide definitive advice.
Robt.
Yes, that's right Shuyi. As I mentioned up above, a Cronbach's alpha can easily be low if you have only a small number of items. (People tend to "worship" Cronbach's alpha far too much; it has largely been discredited - though researchers are resistant to acknowledging that.)
If your two items are reasonably highly correlated with each other (not so much that the are essentially redundant) and form a separate factor, there is some reason to combine them into a single variable.
Robt.
- Badges
- Science topic
- Similar topics
- Phytochemistry
- Extracts