Factor Analysis Using SPSS on a Mainframe Computer
Jo Bell Whitlatch
This is a rather basic inquiry but I would certainly appreciate your help!
The data collected in our study has graduate students posing as users who
ask questions of reference librarians and observe their behavior. After
the interaction the graduate students complete a questionnaire with 45
statements related to librarian behavior. An example is: "The librarian
focuses his/her attention on me." Responses are on a 7-point Likert
scale (strongly disagree --- strongly agree). We also have a series of
questions related to outcome variables with scales ranging from totally
successful to totally unsuccessful.
We wanted to reduce the 45 statements related to librarian behaviors to a
more manageable set and determine underlying concepts. We first ran
factor analysis on 8/10/97 and obtained 10 factors, using mean
substitution for the missing variable treatment, principal components
analysis and varimax rotation. However, using the same set of data on
1/25/98 I ran factor analysis again, using the same defaults, including
mean substitution, but this time received warning message #11302 "the
correlation matrix is ill-conditioned." Has SPSS made some program
changes between the first and second time I ran factor analysis, and
received identical results, except for the warning message? How serious
is this problem? What do you recommend I do?
Once we resolve this problem, we would like to explore the association
between the factors and the outcome variables (i.e. between user
perceptions of librarian behavior and user perceptions of success). Is
correlation (factors with outcome variables) the best SPSS procedure to
use or are there other data analysis methods we should explore?
Thank you for whatever advice you can provide.
Jo Bell Whitlatch
San Jose State University Library
San Jose CA 95192-0028
FAX: 408-924-2701
Email: jwhi...@email.sjsu.edu
Phone: 408-924-2742
Nichols, David
This actually sounds like you were going backwards in terms
of versions. This would make sense if the run on 8/10/97
were on a PC and the later one were on a mainframe. In any
event, you can ignore the ill-conditioned matrix message
if there are no others connected to it. It just means that
the determinant of the matrix is small, but that's not a
big deal.
David Nichols
Principal Support Statistician and
Manager of Statistical Support
SPSS Inc.
nic...@spss.com
>----------
>From: Jo Bell Whitlatch[SMTP:jwhi...@EMAIL.SJSU.EDU]
>Sent: Tuesday, March 03, 1998 12:18 PM
>To: SPS...@UGA.CC.UGA.EDU
>Subject: Factor Analysis Using SPSS on a Mainframe Computer