Due Date: 10/13/05
PM 536: Lab 6
Factor Analysis
This lab exercise is designed to build on your scale construction techniques.
These techniques enable you to assess the validity and reliability of your
measures and to construct more valid and reliable measures. Having better
measures improves the inferences you make from the impact analysis. Use the
data set “labdat_1” that you used for Lab 3 for the following exercises.
Create a log file in which to save your work.
1. Factor q1 to q14 and set the minimum eigenvalue criterion to 1.
Report the number of factors and which items correspond to which factor.
(factor q1-q14, mineigen(1))
2. Measure the reliability with Cronbach’s alpha for the entire scale (Q1-Q14)
by reporting Cronbach’s alpha.
(alpha q1-q14)
3. Create 2 attitude variables from the factor results by summing the items
that load highest on each factor. Report the mean and standard deviation
for the 2 variables. Are they correlated?
(Hint: Use ‘egen” and the “rsum” function)
4. Create 2 attitude variables from the factor results by weighting the new
variables by the factor scores. This is done in a two-step procedure in
SAS. The first step involves running factor analysis with the PROC
FACTOR statement, and outputting the coefficients from the factor pattern
into a new dataset called “fact 1.” The second step involves multiplying
the q1-q16 coefficients from the fact1 dataset with the actual values of q1-
q16 in the original dataset. The resulting scores on each of the factors
can be seen in the new dataset, “scores.” Report the mean and standard
deviation for the 2 variables.
5. Report Cronbach’s alpha for the subscales. Are there any items that can
be dropped to improve the reliabilities?
