If you are going to multiple newsgroups, please send to all of them in
one message so that replies are also sent to all of them.
I don't have a citation at hand.
There has been a long history on the number of factors to retain in
factor analysis. (Google stopping rules )
A lot depends on what what you are doing factor analysis for.
In general you certainly don't want to retain any factor that does not
account for as much variance as an average variable, i.e., that has an
eigenvalue less than 1. This is the Kaiser Criterion. It tells you a
maximum of any plausible number and is (almost) always too high for any
practical purpose. It mainly is used to limit the use of computer time
in extracting useless factors.
There are several other approaches that try to guess how many factors
account for more variance than many factor analyses on data sets with
the same number of variables and cases composed completely of random
data. Google parallel analysis. Google scree test.
These give you ballpark estimates of the number of factors to retain.
You then INTERPRET the factor solution using the number estimated, one
or two more, and smaller numbers of factors.
The number actually retained depend on interpretation of the solutions.
The interpretation is based on the substantive area you are working in
and what you are trying to achieve. For example, in producing summative
scales, you usually want to find at least 4 variables that load cleanly
and with sufficiently high loadings on each factor retained.
Art Kendall
Social Research Consultants
godrobert101@[EMAIL PROTECTED]
wrote:
> My teacher ask me a question. It question is about factor analysis
> when we use. How many components are good. I study many book but can't
> find the answer. If you have any idea, please tell me, if you find any
> thesis or journal about this question is best, thanks!
|
