Class Exercice on Factor Analysis
PM 536/ October 6, 2005
In a recent study, Pentz, Valente and colleagues have investigator how community coalitions
function to deliver evidence based substance abuse prevention programs at the community level.
Part of the survey administered to over 600 community leaders in 24 communities assessed
various aspects of coalition function and performance. Below are 2 scales from the survey and
the results of the factor analysis are on the following pages. Please answer the following
questions:
Questions Answers
Do any items need to
be reverse coded? If so,
which ones?
What are values for the
eigenvalues?
How many variables/
scales should be
constructed from these
data and why?
What variables should
be combined and why?
How would you create
new variables?
What factor loading
threshold seems
reasonable?
Label the new
variables:
How well does Q15c
load on factor 1?
How well does Q15c
load on factor 2?
Which items are most
indicative of factors 1
and 2 and why?
Which items should be
dropped?
What effect does
missing data have on
this analysis?
(1)
