|
⇤ ← Revision 1 as of 2025-10-24 18:15:44
Size: 1551
Comment: Initial commit
|
← Revision 2 as of 2025-10-24 21:03:10 ⇥
Size: 1971
Comment: More clarifications
|
| Deletions are marked like this. | Additions are marked like this. |
| Line 18: | Line 18: |
| }}} | |
| Line 20: | Line 19: |
| {{{ | |
| Line 29: | Line 27: |
| }}} | |
| Line 31: | Line 28: |
| {{{ | |
| Line 36: | Line 32: |
| }}} | |
| Line 38: | Line 33: |
| {{{ | |
| Line 43: | Line 37: |
| }}} Intercept terms are implicitly inserted. The following two are equivalent. {{{ gsem (x1 <- X) (x2 <- X) (x3 <- X) (x4 <- X) gsem (x1 <- X _cons) (x2 <- X _cons) (x3 <- X _cons) (x4 <- X _cons) |
|
| Line 77: | Line 79: |
| ---- == Intercepts == Intercepts are freely estimated by default, but a common design choice is constraining them to 0. This is done like: {{{ gsem (x1 <- X _cons@0) (x2 <- X _cons@0) (x3 <- X _cons@0) (x4 <- X _cons@0) }}} |
Stata SEM Path Specification
Stata has a custom syntax for specifying SEM paths.
Contents
Default and Implicit Behaviors
The path interpretter aggregates paths. The following two are equivalent.
... (X -> x1 x2 x3 x4 x5) ///
(Y -> x1 x2 x3 x6 x7)
... (X -> x1) (X -> x2) (X -> x3) (X -> x4) (X -> x5) ///
(Y -> x1) (Y -> x2) (Y -> x3) (Y -> x6) (Y -> x7)Error variables are implicitly inserted for all endogenous variables, with the loading fixed at 1. The following three are equivalent.
... (x1 <- X) (x2 <- X) (x3 <- X) (x4 <- X)
... (x1 <- X e.x1) ///
(x2 <- X e.x2) ///
(x3 <- X e.x3) ///
(x4 <- X e.x4)
... (x1 <- X e.x1@1) ///
(x2 <- X e.x2@1) ///
(x3 <- X e.x3@1) ///
(x4 <- X e.x4@1)Intercept terms are implicitly inserted. The following two are equivalent.
gsem (x1 <- X) (x2 <- X) (x3 <- X) (x4 <- X) gsem (x1 <- X _cons) (x2 <- X _cons) (x3 <- X _cons) (x4 <- X _cons)
Fixing Parameters
To fix a loading to e.g. 1, try:
... (x1 <- X@1) (x2 <- X) (x3 <- X)
To fix two more more loadings as equal, try:
... (x1 <- X@b) (x2 <- X@b) (x3 <- X)
In this example, b is implicitly declared as a coefficient that is equal across references. Note though that if b does not appear anywhere else, it is effectively unconstrained. Any valid label can be used in place of b.
Regressions
By default, the paths describe linear regressions. -gsem- adds support for other types of regression.
gsem (Y -> x1 x2 x3, oprobit)
Intercepts
Intercepts are freely estimated by default, but a common design choice is constraining them to 0. This is done like:
gsem (x1 <- X _cons@0) (x2 <- X _cons@0) (x3 <- X _cons@0) (x4 <- X _cons@0)