EFFPART effect <, effect …>;
where effect represents:var-list < direction var-list2> and direction is the direction of the effect, as indicated by one of the following: ===>
, --->
, ==>
, -->
, =>
, ->
, >
, <===
, <---
, <==
, <--
, <=
, <-
, or <
.
In the EFFPART statement, you select those effects you want to analyze by partitioning the total effects into direct and indirect effects, with estimated standard errors. The EFFPART or TOTEFF option of the PROC CALIS statement also enables you to analyze effects. The difference is that the EFFPART or TOTEFF option displays effects on all endogenous variables, while the EFFPART statement shows only the effects of interest. In addition, the EFFPART statement enables you to arrange the effects in any way you like. Hence, the EFFPART statement offers a more precise and organized way to present various results of effects.
The EFFPART statement supports the following three types of effect specifications:
>
, =>
, ->
, ==>
, -->
, ===>
, or --->
direction
Example:
effpart X1 X3-X5 ===> Y1 Y2;
This will display four separate tables, respectively for the effects of X1
, X3
, X4
, and X5
on Y1
and Y2
. Each table contains the total, direct, and indirect effects of an X
-variable on the two Y
-variables.
<
, <=
, <-
, <==
, <--
, <===
, or <---
direction
Example:
effpart Y1 Y2 <=== X1 X3-X5;
This will display two separate tables, respectively for the effects on Y1
and Y2
, by X1
, X3
, X4
, and X5
. Each table contains the total, direct, and indirect effects of the four X
-variables on a Y
-variable. Certainly, the results produced from this statement are essentially the same as the previous statement. The difference
is about the organization of the effects in the tables.
no direction
Example:
effpart Y1 Y2 X1-X3;
In this case, variables on the list are analyzed one by one to determine the nature of the effects. If a variable has nonzero effects on any other variables in the model, a table of the total, direct, and indirect effects of the variable on those variables is displayed. If a variable is endogenous, a table of total, direct, and indirect effects of those variables that have nonzero effects on the variable is displayed. Note that an endogenous variable in a model might also have effects on other endogenous variables. Therefore, the two cases mentioned are not mutually exclusive—a variable listed in the EFFPART statement might yield two tables for effect analysis.