In this example, the
variable of interest is
Age at Death, which
should be the first variable listed in the
Measure section
of the
Data pane. Because you want this variable to be the
response variable, click, drag, and drop
Age
at Death from the
Data pane
onto the model pane. Notice that
Age at Death now
appears in the
Response field on the
Roles tab.
The next step is to choose the effect variables or interaction terms that you want
to include in the analysis. One option is to make every available variable an effect
variable and let SAS Visual Statistics perform
variable selection. However, this is not always feasible from a
computational resources perspective. This example creates an interaction term to use as an effect
variable and includes a few other variables as effect variables.
Because you suspect
that systolic blood pressure and diastolic blood pressure interact
with each other, create an interaction term for these variables. In
the
Data pane, click the
icon, and select
New Interaction Effect.
In the
New Interaction Effect window, move
Diastolic and
Systolic from
the
Available columns area into the
Effect
elements area. Click
Create.
The interaction term
Diastolic*Systolic appears
in the
Interaction Effects group of the
Data pane.
Click, drag, and drop
Diastolic*Systolic onto
the model pane. A model is created based on that single effect because
the
Auto-update model option is selected in the right pane. Each time a change is made to the model, the
Linear Regression automatically updates. If you anticipate making many changes or
if you are experiencing server performance issues,
deselect the
Auto-update
model option. When auto-updates are disabled, you must
click
Update in the right pane to update
the model.
Next, add more effects
to the model. Hold down the Ctrl key, and select Blood
Pressure Status, Cause of Death, Leaf
ID 1, Sex, Smoking
Status, Cholesterol, Height, Smoking,
and Weight. Drag and drop these variables
onto the model pane. The Linear Regression updates to include these
effects.
In the right pane, select
the
Properties tab. In this model,
Informative
missingness and
Use variable selection are
not selected. Disabling
Informative missingness means that
observations with
missing values are not included in the analysis. Disabling
Use variable selection means that
all variables are used in the model, regardless of how significant
they are to the model. For this model, keep the default properties
settings.
The Fit
Summary window indicates that Cause of Death, Leaf
ID (1), and Height are the three
most important effects in this model.
The
Assessment window indicates that the observed average and predicted average are approximately
equal for most
bins.
Save the project.