Data
Reduction procedures include
Factor analysis, Correspondence Analysis and Optimal Scaling.
Factor Analysis enables us to do Factor
Analysis. This procedure is for taking a large number of
variables and reducing them into a small number of factors that explain most of
the variance that is observed with the large number of variables.
|
Selection variable: Allows you to limit the analysis to certain levels of
a variable. |
The “Factor Analysis” dialog box has the following submenus:
|
Descriptives- This is chosen if you want
descriptive statistics for all the chosen variables. One can also obtain the
correlation matrix. |
|
|
Extraction- This is where you choose the method you want to use for the
extraction. Principal components (the default) is the method most often
used. You can either choose to extract eigenvalues over a specified
value, or you can choose the number of factors
to extract. |
|
|
Rotation- You choose rotation to make the results easier to
interpret. The default rotation is none, so if you want this feature,
you must choose it. Varimax is the rotation method that is most popular,
although there are others. |
|
|
Scores- If you would like to save the scores obtained by the factor
analysis, you must choose the scores subdialog box
and check “Save as variables”. |
|
|
Options- This is chosen to control how the coefficients are displayed and
how to handle missing values. |
The data used for this demonstration is the Tech Survey data set.
See Data Set page for details. The purpose is
to determine a small set of factors that represent as much information as the
entire 12 variables from question Q31a1 through question Q31a12. These
questions deal with the level of difficulty by faculty when they use technology
in their classrooms. These questions are rated on a scale of 1 to 5, where 1 is
the least difficult to use and 5 is the most difficult to use. We would like to
see if we can reduce these 12 variables into a smaller number of factors.