Factor analysis of multiple measurememts per subject

[rubentech]

Hi
I am new to SPSS, and interested to carry out a factor analysis on my
instrument.I have repeated measurements for each subject (Total
N=145).
When I ran a PCA, this is the result I got with a warning.

"There are fewer than two cases, at least one of the variables has
zero variance, there is only one variable in the analysis, or
correlation coefficients could not be computed for all pairs of
variables. No further statistics will be computed".

I have lots of missing data as well, do I opt for listwise, pairwise
deletion?
Please help!
Thanks a lot.
Ruben

[Art Kendall]

1) Are you sure you want to do a factor analysis and not factorial
analysis of variance?
2) Are you dealing with items from scales?
3) Do you perhaps have dichotomous items with not endorsed set to
missing? to check this go to the variables view

4) about the missing data, to check this, although the descriptive stats
and correlations are available, it is often easier to deal with the
correlation procedure.
correlations variables = var01 to var20
/missing = pairwise
/statistics = descriptives.
correlations variables = var01 to var20
/missing = listwise
/statistics = descriptives.

Hope this helps.

Art
A...@DrKendall.org
Social Research Consultants
University Park, MD USA
(301) 864-5570

[Richard Ulrich]

On 30 Apr 2004 14:08:34 -0700, rubent...@yahoo.com (rubentech)
wrote:

As your final statement suggests -- Yes, the default is
listwise deletion, so every case was dropped whenever
a single variable was missing.

But it is better to clean up your data than do analyze 'pairwise'.
"Lots of missing data" on any variable makes it a poor
candidate for an additive factor. What are you going to
add in, when you come to the ultimate step of trying to score
up the factor? -- If there is a reasonable score to replace
a particular Missing, you want to Recode to that score, prior
to trying a factor analysis....

IS there a variable with zero variance? Drop it, it cannot
add anything to any analysis or conclusions. Drop any
variable where there are a lot of missing that you can't
recode. At this point: How many variables do you have?

Art's post shows the commands for using Correlation to
give you a closer look at what is left.

The fact that you report a lot of missing makes me wonder,
What makes these data a candidate for factor analysis, or
for Principal Components Analysis in particular?
If you want to decompose your data with PCA, then the
fruitful method might be to (a) append a set of
'missing-indicator' variables, (b) set each missing to the mean,
and (c) analyze the doubled set of variables (one extra variable
for each variable that has any missing values)..

--
Rich Ulrich, wpi...@pitt.edu
http://www.pitt.edu/~wpilib/index.html